This page intentionally left blank
DARK ENERGY Theory and Observations Dark energy, the mysterious cause of the accel...
26 downloads
1187 Views
3MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
This page intentionally left blank
DARK ENERGY Theory and Observations Dark energy, the mysterious cause of the accelerating expansion of the Universe, is one of the most important fields of research in astrophysics and cosmology today. Introducing the theoretical ideas, observational methods and results, this textbook is ideally suited to graduate courses on dark energy, and will also supplement advanced cosmology courses. Providing a thorough introduction to this exciting field, the textbook covers the cosmological constant, quintessence, k-essence, perfect fluid models, extradimensional models, and modified gravity. Observational research is reviewed, from the Cosmic Microwave Background to baryon acoustic oscillations, weak lensing, and cluster abundances. Every chapter ends with problems, with full solutions provided, and any calculations are worked through step-by-step. luca amendola is Professor of Physics at the University of Heidelberg, Germany. He is also an astronomer at the Italian National Institute for Astrophysics, Rome, and has authored more than 100 papers in international journals and conference proceedings. shinji tsujikawa is Associate Professor at Tokyo University of Science in Japan, where he teaches cosmology and relativity. He has published more than 80 papers in international refereed journals and has over 60 collaborators worldwide.
DARK ENERGY Theory and Observations LUCA AMENDOLA Institut f¨ur Theoretische Physik, Universit¨at Heidelberg/ Osservatorio Astronomico di Roma
SHINJI TSUJIKAWA Department of Physics, Faculty of Science, Tokyo University of Science
CAMBRIDGE UNIVERSITY PRESS
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo, Delhi, Dubai, Tokyo Cambridge University Press The Edinburgh Building, Cambridge CB2 8RU, UK Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521516006 © L. Amendola and S. Tsujikawa 2010 This publication is in copyright. Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published in print format 2010 ISBN-13
978-0-511-75008-3
eBook (NetLibrary)
ISBN-13
978-0-521-51600-6
Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of urls for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
Contents
1 2
3
4
Preface List of frequently used symbols Overview Expansion history of the Universe 2.1 Friedmann equations 2.2 Hubble’s law 2.3 Matter species in the Universe 2.4 Cosmic distances 2.5 The equation of state of dark energy 2.6 Problems Correlation function and power spectrum 3.1 The correlation function 3.2 The n-point correlation function 3.3 The power spectrum 3.4 From the power spectrum to the moments 3.5 Problems Basics of cosmological perturbation theory 4.1 Perturbing General Relativity 4.2 The Newtonian gauge 4.3 Single-fluid model 4.4 Scales larger than the horizon 4.5 Scales smaller than the Hubble radius 4.6 Two-fluid solutions 4.7 Velocity field 4.8 The redshift distortion 4.9 Baryons, photons, and neutrinos 4.10 The matter power spectrum
v
page ix xiii 1 7 7 12 13 18 22 26 27 27 30 31 37 39 40 40 42 46 51 51 53 56 58 62 70
vi
5
6
7
8
9
Contents
4.11 Perturbed photon propagation 4.12 Problems Observational evidence of dark energy 5.1 The age of the Universe 5.2 Supernova observations 5.3 Cosmic Microwave Background 5.4 Baryon acoustic oscillations 5.5 Large-scale structure 5.6 Problems Cosmological constant 6.1 Einstein equations with the cosmological constant 6.2 History of the cosmological constant 6.3 The fine tuning problem 6.4 The coincidence problem 6.5 Supersymmetric models 6.6 Cosmological constant and the anthropic principle 6.7 The decoupling of the cosmological constant from gravity 6.8 Problems Dark energy as a modified form of matter I: Quintessence 7.1 Quintessence 7.2 Dynamical system approach 7.3 Early dark energy 7.4 Quintessence potentials in particle physics 7.5 Reconstruction of quintessence from observations 7.6 Problems Dark energy as a modified form of matter II 8.1 k-essence 8.2 Phantoms 8.3 Coupled dark energy 8.4 Chameleon scalar fields 8.5 Dark energy models with scaling solutions 8.6 Unified models of dark energy and dark matter 8.7 Future singularities 8.8 Problems Dark energy as a modification of gravity 9.1 f (R) gravity 9.2 Scalar-tensor theories 9.3 Gauss–Bonnet dark energy models 9.4 Braneworld models of dark energy 9.5 Problems
76 83 84 84 87 93 102 106 108 109 110 111 113 114 116 124 129 133 134 135 138 149 153 163 171 172 172 186 189 205 215 225 230 233 234 234 257 269 277 283
Contents
10 Cosmic acceleration without dark energy 10.1 Void models 10.2 Backreaction 10.3 Problems 11 Dark energy and linear cosmological perturbations 11.1 Perturbations in a general dark energy cosmology 11.2 Perturbations of a scalar field 11.3 From dark energy to dark force 11.4 A massive dark energy field 11.5 Sound speed of a scalar field 11.6 Perturbations in modified gravity models 11.7 Problems 12 Non-linear cosmological perturbations 12.1 Second-order perturbations 12.2 The bispectrum and the higher-order correction to the power spectrum 12.3 Spherical collapse 12.4 The mass function of collapsed objects 12.5 Dark energy N -body simulations 12.6 Problems 13 Statistical methods in cosmology 13.1 The likelihood function 13.2 Model selection 13.3 Fisher matrix 13.4 The Fisher matrix for the power spectrum 13.5 Principal component analysis 13.6 Problems 14 Future observational constraints on the nature of dark energy 14.1 Dark energy and the CMB 14.2 Large-scale structure 14.3 Growth function 14.4 Cosmic shear 14.5 Cluster abundances and baryon fraction 14.6 Other probes 14.7 Problems 15 Conclusion and outlook 16 Answers to the problems 17 Mathematical Appendix References Index
vii
285 286 292 295 296 296 306 309 313 314 317 335 336 336 341 347 351 353 355 356 356 363 367 376 379 381 383 383 391 398 403 409 415 426 427 430 455 457 485
Preface
Perhaps the first recognition that the matter composing the universe may be different from the one we touch and experience every day has been put in writing by early Greek philosophers and by Aristotle in particular. In his work On the Heavens, Aristotle argues that the nature and movement of the stars and planets is so fundamentally different from Earth-like elements that a new substance is required, a “bodily substance other than the formations we know, prior to them all and more divine than they.”1 Later on this cosmic element came to be called quintessentia, or fifth element, and drawing on Plato’s classification of the elements a dodecahedron’s figure was associated with it. More than two thousand years later, astrophysicists have begun to pile up evidence that a new form of matter pervades our Universe. This idea is based on observations that reminds one of Aristotle’s thoughts: the global movement we observe in distant reaches of our cosmos is unexplainable by ordinary matter. All the matter we see on Earth, in the solar system, inside our Galaxy or in similar structures across the Universe has a small or negligible positive pressure and clumps under the influence of gravity. An expanding Universe filled with this form of matter would by necessity slow down. But in 1998, astronomers studying the global expansion by the use of supernovae found that their observed luminosities can be explained only by an accelerated expansion of the Universe. After a full decade of more observations, more analyses and more interpretations, we still cannot find a better explanation than invoking something new, a new force or a new matter, that acts “on the heavens.” This new form of matter, called dark energy, is what this book is about. We do not know the nature of dark energy yet. We are beginning to characterize its properties in several ways, from its abundance to its dynamics, but we know still very little about it. The simplest explanation, an energy associated with the 1
Aristole, On the Heavens, I, 2.
ix
x
Preface
vacuum, was already proposed on a totally different basis by Einstein under the name of universal (Einstein’s original term) or cosmological constant2 . But the freedom still allowed by observations has unleashed theorists’ imagination and many interpretations of dark energy have been advanced. Dark energy is indeed a general label for what we do not know about the large-scale properties of our Universe, its history and its geography. So why a book on dark energy? One of the immediate consequences of the discovery of the cosmic acceleration and the hypothesis of dark energy has been that astrophysicists and particle physicists, both theorists and experimentalists, have been drawn together into this new field of research, with their own languages and methods. We believe that this has created the need for a resource that allows scholars and students to apprehend the basis of dark energy research in an interdisciplinary way. This book introduces the main theoretical ideas on dark energy and at the same time the basics of the observational methods and results. There are several reviews that cover parts of the dark energy research but not a book that could be used as a starting point to advanced and more topical material. This book can be used as a companion text for an advanced cosmology course, covering several areas that complement modern cosmology textbooks or as a standalone text for graduate or post-graduate courses on dark energy. It is also addressed to newcomers in the field that wish to identify the main lines of the current research. Finally, we have in mind also researchers in the dark energy field who need to explore other sides of the discipline and would like to have a handy reference for many results and topics scattered in the literature. For most of the book we assume knowledge of General Relativity and basic cosmology at the graduate level and little more. Some more advanced sections (especially Sections 6.5 and 7.4) require also a background in quantum field theory but they can be left aside without prejudice. Whenever possible we give a fairly comprehensive review of the tools required for further material, for instance we introduce the concepts of statistics and of cosmological perturbation theory that are needed for understanding the subsequent chapters. We provide 44 fully solved problems with some detailed calculations, which will help the reader to test his/her understanding. The immense impact on cosmology of the dark energy concept is witnessed by the many projects around the world aiming at collecting more and more data, from large-scale galaxy surveys to weak lensing surveys, from cosmic microwave observations to gamma-ray bursts. The very nature of the issue at stake, the study of a component that determines to a large extent the present and future cosmic dynamics, has generated a great diversity of theoretical and observational approaches. One 2
As a curious coincidence, Aristotle first proposed his eternal and incorruptible “cosmic substance” in the book (i.e. the twelfth book) of the Metaphysics. “The Lambda Book,” as it was called in the middle ages, was singled out as the highest point of Aristotle’s metaphysics.
Preface
xi
can attack the problem from the point of view of an exotic matter component, or of a non-Einsteinian gravity force, or invoking multidimensional effects. Similarly, one can employ a very diverse array of observations, from standard candles to standard clocks, from supernovae to quasar multiple images. Many of the hypotheses and methods will have a strong impact on cosmology even beyond the dark energy problem. This book tries to extract from this variety the core teachings: methods, suggestions, hypotheses, and techniques that are shaping our knowledge of the cosmos. Many of these, we reckon, will remain with us for many years. Although the diversity of approaches is one of the hallmarks of dark energy research, we could not possibly cover all the ideas discussed so far. Up to 2009 the number of papers that include the words “dark energy” or “cosmological constant” in the title has been over 3700. We tried to discuss all the driving ideas but not all possible implementations. In doing so we certainly missed some interesting contributions; we apologize in advance to our colleagues. We thank all our collaborators on the topics in this book, namely, Carlo Baccigalupi, Amedeo Balbi, Marco Baldi, Kazuharu Bamba, Riccardo Barbieri, Bruce A. Bassett, Silvio Bonometto, Stefano Borgani, Robert Brandenberger, Carlo Burigana, Paolo Cabella, Gianluca Calcagni, Gabriela C. Campos, Salvatore Capozziello, Daniela Carturan, Christos Charmousis, Edmund J. Copeland, Pier Stefano Corasaniti, Stephen C. Davis, Antonio De Felice, Cinzia Di Porto, Stephane Fay, Fabio Finelli, Radouane Gannouji, Mohammad Reza Garousi, Maurizio Gasperini, Chao-Qiang Geng, Emanuele Giallongo, Fabio Giovi, Burin Gumjudpai, Zong-Kuan Guo, Soo A. Kim, Martin Kunz, Maxim Libanov, Andew R. Liddle, Roy Maartens, Andrea Macci`o, Kei-ichi Maeda, Roberto Mainini, Elisabetta Majerotto, Martin Makler, Matteo Martinelli, Alessandro Melchiorri, Shuntaro Mizuno, Bruno Moraes, David F. Mota, Tapan Naskar, Savvas Nesseris, Shin’ichi Nojiri, Sergei Odintsov, Junko Ohashi, Nobuyoshi Ohta, Sudhakar Panda, Eleftherios Papantonopoulos, David Parkinson, Alessandro Pasqui, Valeria Pettorino, Yun-Song Piao, Federico Piazza, David Polarski, Miguel Quartin, Claudia Quercellini, R. R. R. Reis, Rogerio Rosenfeld, Valery Rubakov, M. Sami, Domenico Sapone, Parampreet Singh, Alexei Starobinsky, Takashi Tamaki, Takayuki Tatekawa, Reza Tavakol, Domenico Tocchini-Valentini, Alexey Toporensky, Peter V. Tretjakov, Roberto Trotta, Kotub Uddin, Carlo Ungarelli, Ioav Waga, David Wands, John Ward, Christof Wetterich, Jun’ichi Yokoyama, and Xinmin Zhang. Apart from collaborators, we are grateful to many other people with whom we have discussed a lot about dark energy. We also thank Graham Hart and the editorial staff at Cambridge University Press for giving us the opportunity to write this book. Finally we are grateful to our families: Emanuela, Davide, Yasuko, and Masato for their support and love. L.A. and S.T.
xii
Preface
I wish to honor the memory of Franco Occhionero, long-time friend and teacher: you opened to me the magic lamp of cosmology. I also thank the staff and colleagues at the Osservatorio Astronomico di Roma and the Director Emanuele Giallongo, for support, patience, collaboration, help, friendship, and everything that makes work possible, pleasant and productive. Thanks to the University of Heidelberg for the new exciting opportunities they are offering to me. Special thanks to Silvio Bonometto, Roberto Buonanno, Roberto Scaramella and Christof Wetterich for collaboration, advice, support, and friendship. Special thanks to Claudia Quercellini, Ioav Waga, Miguel Quartin, Fabio Finelli, Cinzia di Porto for comments on the draft and for the long-term friendship and collaboration. Thanks also to Adam Amara, Enzo Branchini, Francisco Castander, Gigi Guzzo, Tom Kitching, Rocky Kolb, Martin Kunz, Roberto Maoli, Anais Rassat, Alexandre Refregier, Yun Wang, Jochen Weller, and all the other DUNE/Euclid and NASA/SWG collaborators, too many to be listed here, from whom I learned a lot in terms both of science and of scientific collaboration. L.A. I am grateful to Kei-ichi Maeda for giving me a chance to work as a cosmologist, in spite of the fact that I was majoring mathematics before entering a master course of physics in 1996. This was a correct decision because I witnessed enormous progress of observational and theoretical cosmology from the late 1990s, especially in the field of dark energy. I am also thankful to the members of Tokyo University of Science and Gunma National College of Technology, especially to Antonio De Felice, Hitoshi Fujiwara and Junko Ohashi, for their kind support and help. I also thank Bruce A. Bassett and Roy Maartens, who kindly helped me in many aspects during my stay in Portsmouth in 2003. I am grateful to M. Sami for a long-term collaboration about dark energy. S.T.
Frequently used symbols
Symbol
Definition
G mpl Mpl κ a t η N ˙
−8 3 −1 Gravitational constant (G g sec−2 ) √ = 6.67 × 10 cm 19 Planck mass (mpl = 1/ G = 1.2211 √ × 10 GeV) Reduced Planck mass (Mpl = 1/ 8π G = 2.4357 × 1018 GeV) √ 8π G Scale factor of the Universe (with the present value a0 = 1) Cosmic time Conformal time: η = a −1 dt Number of e-foldings: N = ln a Derivative with respect to t Derivative with respect to η (or N = ln a in Chapters 11 and 12) Redshift: z = a0 /a − 1 Angular diameter distance, luminosity distance ˙ Hubble parameter: H = a/a, conformal Hubble parameter H = aH Present Hubble parameter: H0 = 100 h km sec−1 Mpc−1 Hubble parameter normalized by H0 : E(z) = H (z)/H0 (Energy) Density Pressure Equation of state: w = P /ρ Effective or total equation of state: weff = −1 − 2H˙ /(3H 2 ) Curvature of the Universe Ricci scalar Density parameter at the present epoch (z = 0) Sound speed η Sound horizon: rs (η) = 0 dη˜ cs (η) ˜ Gravitational potentials Temperature
z dA , dL H, H H0 , h E(z) ρ P w weff K R (0) cs rs , T
xiii
xiv
Frequently used symbols
Symbol
Definition
k P (k) C R δ D b S gµν Gµν Tµν φ V (φ) U (φ) L λ λˆ X Q ωBD 2 RGB
Temperature perturbations: = δT /T Comoving wavenumber Power spectrum of perturbations Spherical harmonic multipoles Multipole power spectrum CMB shift parameter Density contrast Growth function Bias (ratio of galaxy to total matter perturbations) Cosmological constant Action Metric Einstein tensor Energy-momentum tensor Scalar field Scalar-field potential in the Einstein frame Scalar-field potential in the Jordan frame Lagrangian density (also log-likelihood) Slope of the potential defined by λ = −V,φ /(κV ) Dimensionless perturbation scale λˆ = H/k Kinetic energy: X = −(1/2)g µν ∂µ φ∂ν φ Coupling between a scalar field φ and non-relativistic matter Brans–Dicke parameter Gauss–Bonnet term
1 Overview
From the observational data of Supernovae Type Ia (SN Ia) accumulated by the year 1998, Riess et al. [1] in the High-redshift Supernova Search Team and Perlmutter et al. [2] in the Supernova Cosmology Project Team independently reported that the present Universe is accelerating. The source for this late-time cosmic acceleration was dubbed “dark energy.” Despite many years of research (see e.g., the reviews [3, 4, 5, 6, 7]) its origin has not been identified yet. Dark energy is distinguished from ordinary matter species such as baryons and radiation, in the sense that it has a negative pressure. This negative pressure leads to the accelerated expansion of the Universe by counteracting the gravitational force. The SN Ia observations have shown that about 70% of the present energy of the Universe consists of dark energy. The expression “dark energy” may be somewhat confusing in the sense that a similar expression, “dark matter,” has been used to describe a pressureless matter (a non-relativistic matter) that interacts very weakly with standard matter particles. The existence of dark matter was already pointed out by Zwicky in the 1930s by comparing the dispersion velocities of galaxies in the Coma cluster with the observable star mass. Since dark matter does not mediate the electromagnetic force, its presence is mainly inferred from gravitational effects on visible matter. Dark matter can cluster by gravitational instability (unlike standard dark energy) so that local structures have been formed in the Universe. In fact it is observationally known that dark matter has played a crucial role for the growth of large-scale structure such as galaxies and clusters of galaxies. The energy fraction of dark matter in the present universe is about 25%, whereas that of baryons is about 4%. The black body radiation, which dominated over the other matter components in the past, shares only about 0.005% of the present total energy density. In modern cosmology it is believed that another cosmic acceleration called “inflation” occurred in the very early Universe prior to the radiation-dominated
1
2
Overview
epoch. The idea of inflation was originally proposed in the early 1980s by a number of people [8, 9, 10, 11] to solve several cosmological problems such as the flatness and horizon problems. Inflation also provides a causal mechanism for the origin of large-scale structure in the Universe. The temperature anisotropies of Cosmic Microwave Background (CMB) observed by the Cosmic Background Explorer (COBE) in 1992 [12] showed that the fluctuation spectrum is nearly scale-invariant.1 This is consistent with theoretical predictions of the power spectrum of density perturbations originated from quantum fluctuations of a scalar field generated during inflation. After 2003, the Wilkinson Microwave Anisotropy Probe (WMAP) group has provided high-precision observational data of CMB anisotropies [13, 14, 15]. This has given strong support for the existence of an inflationary period as well as dark energy. After the end of inflation the Universe entered the radiation-dominated epoch during which light elements such as helium and deuterium were formed. Since the energy density of radiation decreases faster than that of non-relativistic matter such as dark matter and baryons, the radiation-dominated era is eventually followed by the matter-dominated epoch around the redshift z = 3000. The temperature anisotropies observed by COBE and WMAP occur on the last scattering surface at which electrons were trapped by hydrogen to form atoms. After this decoupling epoch photons can freely move to us without experiencing Thomson scattering. The decoupling corresponds to the redshift z 1090. According to the WMAP 5-year data [15], the energy components at the decoupling epoch are dark matter (63%), radiations (25%) [photons (15%) and neutrinos (10%)], and baryons (12%) with at most a tiny amount of dark energy. We will often make reference to the cosmological parameters measured by WMAP in the course of this book. The formation of structure (galaxies, clusters) started in the matter-dominated epoch, i.e. when the pressureless dark matter began to dominate the total energy density of the Universe. Baryons also contribute to the formation of large-scale structure to some extent. During the matter era the energy density of dark energy needs to be suppressed compared to that of dark matter in order to allow sufficient growth of large-scale structure. If dark energy couples to dark matter with some interaction (as in the coupled quintessence scenario [16, 17]), then dark energy also affects the past expansion history of the Universe as well as the structure formation. It is possible to place bounds on the strength of such couplings from the observations of CMB and of galaxy clustering. In addition to the experiments of direct and indirect dark matter search (see e.g., [18, 19, 20, 21]), like those at the Large Hadron Collider (LHC) at CERN and in underground, ground, and space 1
J. Mather and G. Smoot won the Nobel Prize in 2006 for the measurement of the black body spectrum and the discovery of the temperature anisotropy of CMB.
Overview
3
facilities, cosmological observations will shed light on the relation between dark matter and dark energy. While the energy density of dark matter evolves as ρm ∝ a −3 (a is the scale factor of an expanding Universe), the dark energy density is nearly constant in time (ρDE ∝ a −n with n probably close to 0). Hence the latter energy density eventually catches up with the former. The onset of the cosmic acceleration occurs around the redshift z ∼ 1, although there is still uncertainty for its precise value because of the model-dependence. We live then in a special epoch of the cosmic acceleration in the long expansion history of the Universe. The problem why the accelerated expansion of the Universe started around today is often called the “coincidence problem.” The standard radiation- and matter-dominated eras are sandwiched by two periods of cosmic acceleration – inflation and dark energy. The simplest candidate for dark energy is the so-called cosmological constant , whose energy density remains constant [22]. Originally the cosmological constant was introduced by Einstein in 1917 to realize a static Universe in the framework of General Relativity [23]. In fact the Einstein equations allow the freedom to add the constant term. The cosmological constant works as a negative pressure against gravity so that the two effects can balance each other. However, after the discovery of the expansion of the Universe by Hubble from the measurement of recession speeds of distant galaxies, Einstein abandoned the idea of adding the term to the Einstein equations. At the late stage of his career, he regretted having introduced as his “biggest blunder” (or so is told by George Gamow). In fact, there was nothing to regret: after 1998 the cosmological constant revived again as a form of dark energy responsible for the late-time acceleration of the Universe. From the viewpoint of particle physics, the cosmological constant appears as vacuum energy density. If we sum up zero-point energies of all normal modes of some field and take the cut-off scale of the momentum at the Planck scale, the vacuum energy density is estimated to be ρvac 1074 GeV4 . This is much larger than the observed value of dark energy: ρ 10−47 GeV4 . If vacuum energy with an energy density of the order of ρvac 1074 GeV4 was present in the past, the Universe would have entered an eternal stage of cosmic acceleration already in the very early Universe. This is of course problematic because the success of the big bang cosmology based on the presence of radiation and matter epochs is completely destroyed. Hence the problem of the large vacuum energy density was known long before the discovery of dark energy in 1998. If the cosmological constant is responsible for the present cosmic acceleration, we need to find a mechanism to obtain the tiny value of consistent with observations. A lot of efforts have been made in this direction under the framework of particle physics. For example, the recent development of string theory shows that it is possible to construct de Sitter vacua by compactifying extra dimensions in the
4
Overview
presence of fluxes with an account of non-perturbative corrections [24]. The fact that there is a huge number of different choices of fluxes gives rise to the so-called “string landscape” with more than 10500 vacua [25]. Some scientists argued that only the vacuum whose energy density is of the order of the present cosmological density can sustain life or complexity and this explains why we live in a low- world. This anthropic argument is, to say the least, highly controversial. If the origin of dark energy is not the cosmological constant, one may seek for some alternative models to explain the cosmic acceleration today. Basically there are two approaches to construct models of dark energy other than the cosmological constant. The first approach is to modify the right-hand side (r.h.s.) of the Einstein equations given in Eq. (2.8) by considering specific forms of the energy-momentum tensor Tµν with a negative pressure. The representative models that belong to this class are the so-called cosmon or quintessence [26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41], k-essence [42, 43, 44], and perfect fluid models [45, 46]. The quintessence makes use of scalar fields with slowly varying potentials, whereas in k-essence it is the scalar-field kinetic energy that drives the acceleration. The perfect fluid models are based on a perfect fluid with a specific equation of state such as the Chaplygin gas model [45] and its generalizations [46]. There have been many attempts to construct scalar-field models of dark energy based on particle physics (see Refs. [47, 48, 49, 50, 51, 52] for early works). In the context of inflation, since the associated energy scale is high, it is natural for scalar fields to be responsible for the acceleration of the Universe. The situation is different for dark energy – its energy scale is too low compared to typical scales appearing in particle physics. Moreover, the field potentials need to be sufficiently flat so that the field evolves slowly enough to drive the present cosmic acceleration. This demands that the field mass is extremely light (mφ 10−33 eV) relative to typical mass scales appearing in particle physics. It would be expected that this light scalar field should mediate long-range forces with ordinary matter [36]. Such couplings need to be suppressed in order to be consistent with a number of local gravity experiments. In spite of the above-mentioned difficulties it is not hopeless to construct viable scalar-field dark energy models in the framework of particle physics. The second approach for the construction of dark energy models is to modify the left-hand side (l.h.s.) of the Einstein equations (2.8). The representative models that belong to this class (that we denote “modified gravity”) are the so-called f (R) gravity [53, 54, 55], scalar-tensor theories [56, 57, 58, 59, 60], and braneworld models [61, 62]. The cosmological constant scenario (in other words, the “-ColdDark-Matter (CDM) model”) corresponds to the Lagrangian density f (R) = R − 2, where R is the Ricci scalar. A possible modification of the CDM is described by a non-linear Lagrangian density f in terms of R, which is called
Overview
5
f (R) gravity. Scalar-tensor theories correspond to theories in which the Ricci scalar R couples to a scalar field φ with a coupling of the form F (φ)R. They include Brans–Dicke theory [63] and dilaton gravity [64] as specific cases. In the braneworld models proposed by Dvali, Gabadadze, and Porrati (DGP) [61] the late-time acceleration of the Universe can be realized as a result of the gravitational leakage from a 3-dimensional surface (3-brane) to a 5-th extra dimension on Hubble distances. Generally we require that modified gravity models satisfy local gravity constraints as well as conditions for the cosmic acceleration preceded by the matter-dominated epoch. In this sense modified gravity models are typically more strongly constrained than modified matter models from gravitational experiments and cosmological observations. It is important to realize however that the two approaches, which we denote as modified matter and modified gravity, are not fundamentally different, at least if for a moment we do not consider their quantum field implications. From the viewpoint of classical General Relativity (which is all that matters for most of cosmology), one can always rephrase one into the other by defining a suitable conserved energy-momentum tensor that equals the Einstein tensor. In order to distinguish this variety of models of dark energy, it is important to place constraints by using observational data such as SN Ia, CMB, and large-scale structure (LSS). Usually the equation of state of dark energy, wDE ≡ PDE /ρDE , where PDE is the pressure and ρDE is the energy density, is a good measure to describe the property of dark energy at background level. In the case of the cosmological constant we have PDE = −ρDE and hence wDE = −1. In other models of dark energy the equation of state wDE generally varies in time. Perhaps the first task of dark energy research is to detect deviations from the value wDE = −1 in order to find whether dark energy can be identified with the cosmological constant or not. The SN Ia observations have provided information of the cosmic expansion history around the redshift z 2 by the measurement of luminosity distances of the sources. The presence of dark energy leads to a shift of the position of acoustic peaks in CMB anisotropies as well as a modification of the large-scale CMB spectrum through the so-called integrated Sachs–Wolfe effect. Although the CMB data alone are not sufficient to place strong constraints on dark energy, the combined analysis of SN Ia and CMB can provide tight bounds on the equation of state wDE and the present energy fraction (0) DE of dark energy [15]. The distribution of large-scale clustering of galaxies in the sky also provides additional information on the properties of dark energy [65, 66, 67]. In 2005 the detection of a peak of baryon acoustic oscillations (BAO) was reported by Eisenstein et al. [68] at the average redshift z = 0.35 from the observations of luminous red galaxies in the Sloan Digital Sky Survey. This has also given us another independent test of dark
6
Overview
energy. From the combined analysis of SN Ia, CMB, and BAO, the WMAP group [15] obtained the bound −1.097 < wDE < −0.858 at the 95% confidence level assuming a constant equation of state. The cosmological constant (wDE = −1) is well consistent with the current observational data while some dark energy models have been already excluded from observations. In future observations it is expected that other observational data such as weak gravitational lensing and gamma ray bursts will shed light on the nature of dark energy. Confirming CDM or detecting deviations from it would be an extremely important step towards understanding the origin of dark energy. Units and conventions Throughout this book we use units such that c = = kB = 1, where c is the speed of light, is reduced Planck’s constant, and kB is Boltzmann’s constant. We reinsert these symbols when the discussion needs it. In these units everything can be expressed in terms of a single unit, e.g., time, length, or energy. The gravitational constant G is related to the Planck mass mpl = 1.2211 × 1019 GeV via G = 1/m2pl and the reduced Planck mass Mpl = 2.4357 × 1018 GeV via κ 2 ≡ 8π G = 1/Mpl2 , respectively. We adopt the metric signature (−, +, +, +). We list frequently used symbols after the Table of Contents.
2 Expansion history of the Universe
Standard hot big bang cosmology is based on the cosmological principle, which states that the Universe is homogeneous and isotropic at least on large scales. This is supported by a number of observations, such as the CMB photons coming from different parts of the sky with almost the same temperature. The past cosmic expansion history is recovered by solving the Einstein equations in the background of the homogeneous and isotropic Universe. Of course we observe inhomogeneities and irregularities in the local region of the Universe such as stars and galaxies. These inhomogeneities have grown in time through gravitational instability from a matter distribution that was more homogeneous in the past. Then the inhomogeneities can be regarded as small perturbations evolving on the background (homogeneous) Universe. In this chapter we provide the basic tools to understand the expansion history of the Universe. We also introduce a number of cosmic distances often used to put observational constraints on dark energy.
2.1 Friedmann equations The line-element that describes a 4-dimensional homogeneous and isotropic spacetime is called Friedmann–Lemaˆıtre–Robertson–Walker (FLRW) spacetime and is given by ds 2 = gµν dx µ dx ν = −dt 2 + a 2 (t)dσ 2 ,
(2.1)
where gµν is a metric tensor, a(t) is a scale factor with cosmic time t, and dσ 2 is the time-independent metric of the 3-dimensional space with a constant curvature K: dσ 2 = γij dx i dx j =
dr 2 + r 2 (dθ 2 + sin2 θ dφ 2 ) . 2 1 − Kr 7
(2.2)
8
Expansion history of the Universe
Here K = +1, −1, 0 correspond to closed, open, and flat geometries, respectively. We have used polar coordinates (x 1 , x 2 , x 3 ) = (r, θ, φ) with γ11 = (1 − Kr 2 )−1 , γ22 = r 2 , and γ33 = r 2 sin2 θ . In Eq. (2.1) the Greek indices µ and ν run from 0 to 3, whereas in Eq. (2.2) the Latin indices i and j run from 1 to 3; the same convention applies to the whole book except when indicated otherwise. We follow Einstein’s convention that the terms with same upper and lower indices are summed over. See the book of Weinberg [69] for the derivation of the metric (2.1) from a maximally symmetric spacetime. In addition to the cosmic time t, we also introduce the conformal time η defined by (2.3) η ≡ a −1 dt . The dynamical equations of motion in the expanding Universe can be derived from the Einstein equations by the following steps. From the metric gµν we obtain the Christoffel symbol: 1 µ νλ = g µα (gαν,λ + gαλ,ν − gνλ,α ) , 2
(2.4)
where gαν,λ ≡ ∂gαν /∂x λ . Note that gαν satisfies the relation g µα gαν = δνµ , where δνµ is Kronecker’s delta (δνµ =1 for µ = ν and δνµ =0 for µ = ν). The Ricci tensor is defined by β
α α α α β − µα,ν + µν αβ − µβ αν . Rµν = µν,α
(2.5)
The contraction of the Ricci tensor gives the Ricci scalar (scalar curvature) R = g µν Rµν .
(2.6)
We can then evaluate the Einstein tensor 1 Gµν ≡ Rµν − gµν R . 2
(2.7)
The cosmological dynamics can be obtained by solving the Einstein equations Gµν = 8π GTνµ ,
(2.8)
where Tνµ is the energy-momentum tensor of matter components. The l.h.s. of Eq. (2.8) characterizes the geometry of spacetime, whereas the r.h.s. describes energies and momenta of matter components. In the cosmological setting the cosmic expansion rate is determined by specifying the properties of matter in the Universe.
2.1 Friedmann equations
9
For the FLRW metric (2.1) the non-vanishing components of Christoffel symbols are i = ji 0 = H δji , ij0 = a 2 H γij , 0j
(2.9)
Kr 1 1 , 22 = −r(1 − Kr 2 ) , 33 = −r(1 − Kr 2 ) sin2 θ, (2.10) 1 − Kr 2 1 2 2 3 3 3 3 = − sin θ cos θ , 12 = 21 = 13 = 31 = , 23 = 32 = cot θ, r (2.11)
1 11 = 2 33
where ˙ H ≡ a/a.
(2.12)
A dot represents a derivative with respect to cosmic time t. The quantity H , called the Hubble parameter, describes the expansion rate of the Universe. The Christoffel symbols given in Eqs. (2.10) and (2.11) correspond to those for the three-dimensional metric (2.2) with the curvature K. From Eqs. (2.5) and (2.6) the Ricci tensor and the scalar curvature are R00 = −3 H 2 + H˙ , R0i = Ri0 = 0 , R = 6 2H 2 + H˙ + K/a 2 .
Rij = a 2 3H 2 + H˙ + 2K/a 2 γij , (2.13) (2.14)
From Eq. (2.7) together with the relation Gµν = g µα Gαν , the Einstein tensor is G00 = −3 H 2 + K/a 2 ,
G0i = Gi0 = 0 ,
Gij = − 3H 2 + 2H˙ + K/a 2 δji . (2.15)
In the FLRW spacetime the energy-momentum tensor of the background matter is restricted to take the perfect fluid form: Tνµ = (ρ + P )uµ uν + P δνµ ,
(2.16)
where uµ = (−1, 0, 0, 0) is the four-velocity of the fluid in comoving coordinates, and ρ and P are functions of t. The (00) and (ij ) components of Tνµ are T00 = −ρ and Tji = P δji . Then ρ and P have the meaning of an energy density and a pressure, respectively. Since we are using the unit c = 1, the density ρ is not particularly distinguished from the energy density ρc2 . From the (00) and (ii) components of the Einstein equations (2.8) we obtain H2 =
K 8π G ρ− 2, 3 a
K 3H 2 + 2H˙ = −8π GP − 2 . a
(2.17) (2.18)
10
Expansion history of the Universe
Eliminating the term K/a 2 gives a¨ 4π G =− (ρ + 3P ) . a 3
(2.19)
Multiplying Eq. (2.17) by a 2 , differentiating and using Eq. (2.19), we find ρ˙ + 3H (ρ + P ) = 0 .
(2.20)
The Einstein tensor satisfies the Bianchi identities ∂Gµν µ α ∇µ Gµν ≡ + αµ Gαν − νµ Gµα = 0 , (2.21) ∂x µ where ∇µ denotes the covariant derivative. Sometimes we use also the symbol “;µ ” to represent the covariant derivative. From the Einstein equations (2.8) it follows that ∇µ Tνµ = 0, which gives the same equation as (2.20) in the FLRW background (see problem 2.1). Hence Eq. (2.20) is called the conservation or continuity equation. Equation (2.17) can be written in the form: M + K = 1 ,
(2.22)
where M ≡
8π Gρ , 3H 2
K ≡ −
K . (aH )2
(2.23)
We often refer to the present values of the density parameters. For relativistic particles, non-relativistic matter, dark energy, and curvature, we have, respectively (0) r =
8π Gρr(0) , 3H02
(0) m =
8π Gρm(0) , 3H02
(0) DE =
(0) 8π GρDE , 3H02
(0) K =−
K . (a0 H0 )2 (2.24)
When we wish to identify the electromagnetic radiation, rather than all the relativistic particles, we use the subscript γ . When we need to distinguish between (cold) dark matter and baryons we use the subscripts c and b, respectively.1 When we refer to the cosmological constant, we also use the subscript ∧ instead of DE. Finally, sometimes we use M to denote a generic matter component. If the expansion of the Universe is decelerated (i.e. a¨ < 0) then the curvature ˙ decreases), apart term |K | continues to increase (because the term aH (= a) from the case where the Universe is exactly flat (K = 0) from the very beginning. The WMAP 5-year data [15] constrain the curvature of the present Universe to be −0.0175 < (0) K < 0.0085 at the 95% confidence level. We need a phase of 1
See Section 2.3 for the definition of cold dark matter.
2.1 Friedmann equations
11
cosmic acceleration (a¨ > 0) to reduce |K | in the past cosmic expansion history, unless the initial state of the Universe is extremely close to the flat one. In order to realize the present level of flatness of the Universe, we require, prior to the radiation-dominated epoch, a phase of cosmic inflation during which the scale factor changes by more than e70 times [70]. Let us consider the case in which the Universe is dominated by a single component with an equation of state defined by w ≡ P /ρ .
(2.25)
If w is a constant, one can analytically find the evolution of ρ and a for the flat Universe (K = 0). Solving Eqs. (2.17) and (2.20) in this case, we obtain the following solutions ρ ∝ a −3(1+w) ,
a ∝ (t − ti )2/(3(1+w)) ,
(2.26)
where ti is a constant. Since from statistical mechanics we know that radiation has the equation of state w = 1/3 (as we will see later), it follows that the cosmic evolution during the radiation-dominated epoch is given by ρ ∝ a −4 and a ∝ (t − ti )1/2 . Non-relativistic matter corresponds to the case with a negligible pressure relative to its energy density, i.e. w 0. Then the evolution during the matterdominated era is given by ρ ∝ a −3 and a ∝ (t − ti )2/3 . In order to give rise to the cosmic acceleration we require a¨ > 0 in Eq. (2.19), i.e. P < −ρ/3
→
w < −1/3 ,
(2.27)
where ρ is assumed to be positive. The fact that the negative pressure leads to the cosmic acceleration may look counter-intuitive. In Newtonian gravity the pressure is related to a force associated with a local potential that depends on the position in space. In the homogeneous and isotropic Universe such a local potential is absent, which means that there is no Newtonian-analog pressure. In other words, the time-dependent pressure P (t) in the FLRW spacetime appears only in General Relativity. The mechanisms that generate this negative pressure and the cosmic acceleration are the main topic of this book. When w = −1, i.e. P = −ρ, it follows from Eq. (2.20) that ρ is a constant. This case corresponds indeed to the so-called cosmological constant. Since H is constant in the flat Universe (K = 0), the scale factor evolves exponentially: a ∝ exp(H t). The cosmological constant cannot be responsible for inflation in the early Universe because otherwise the accelerated expansion would not end. However, it is possible that the cosmological constant is responsible for dark energy because the current cosmic acceleration might indeed continue without end.
12
Expansion history of the Universe
2.2 Hubble’s law In the 1920s Slipher and Hubble found that the observed wavelength λ0 of absorption lines of distant galaxies is larger than the wavelength λ in the rest frame [71]. This is due to the fact that the wavelength is stretched in proportion to the scale factor in an expanding Universe. In order to quantify this effect, we introduce the redshift a0 λ0 −1= − 1, (2.28) z≡ λ a where the present epoch corresponds to z = 0. In the following we take the present scale factor a0 to be unity unless otherwise stated. As we go back to the past, z gets larger. As long as the recessional velocity v of an object is much smaller than the speed of light c we have λ0 (1 + v/c)λ from the Doppler effect, giving z v/c .
(2.29)
In an expanding Universe a physical distance r from an observer (at the origin) to an object is given by r = a(t)x, where x denotes the comoving distance. For objects moving with the Hubble flow, the comoving distance remains constant. Taking the derivative of the equation r = a(t)x with respect to t, we obtain r˙ = H r + a x˙ .
(2.30)
The velocity v H ≡ H r appears because of the presence of the cosmic expansion. On the other hand, the velocity v p ≡ a x˙ , called peculiar velocity, describes the movement of an object with respect to the local Hubble flow. The speed of the object along the direction from the observer to the object is given by v ≡ r˙ · r/r = H r + v p · r/r ,
(2.31)
where r ≡ |r|. In most cases the peculiar velocity of galaxies does not exceed 106 m/s. Under the condition that the term v p · r/r is negligible relative to the term H r, we obtain v H0 r .
(2.32)
Here we have replaced H for the present value H0 , which is justified in small redshift regions (z 1). In 1929, Hubble reported the law (2.32) by plotting the recessional velocity v versus the distance r. His data were scarce, shallow, and noisy, but Hubble concluded correctly that the universe was expanding. The Hubble parameter H0 (Hubble constant) is usually written as H0 = 100 h km sec−1 Mpc−1 = 2.1332h × 10−42 GeV ,
(2.33)
1 Mpc = 3.08568 × 1024 cm = 3.26156 × 106 light years .
(2.34)
where
2.3 Matter species in the Universe
13
Note that h describes the uncertainty on the value H0 . The observations of the Hubble Key Project constrain this value to be [72] h = 0.72 ± 0.08 .
(2.35)
Originally Hubble derived a much larger value, H0 ∼ 500 km sec−1 Mpc−1 , due to the uncertainty of the measurement of distances at that time. We define the Hubble time tH ≡ 1/H0 = 9.78 × 109 h−1 years ,
(2.36)
which is a rough measure of the age of the Universe. The present Hubble radius is defined by c DH ≡ = 2998 h−1 Mpc , (2.37) H0 which corresponds roughly to the largest scale we can observe now. It is also convenient to introduce the critical density 3H02 = 1.88 h2 × 10−29 g cm−3 , ρc(0) ≡ (2.38) 8π G which represents the averaged cosmological density in the Universe today. Note that we have used Eqs. (2.33) and (2.34) together with the value G = 6.67 × 10−8 cm3 g−1 sec−2 to obtain the numerical value in Eq. (2.38). The critical density (2.38) is very small compared to densities in the local structure of the Universe (ρ 5 g/cm3 for Earth and ρ 10−24 g/cm3 for the homogeneous baryon/dark matter density in our Galaxy). An even smaller fraction is responsible for the present accelerated expansion of the Universe. 2.3 Matter species in the Universe Let us consider matter species in the Universe. They are broadly classified into relativistic particles, non-relativistic matter, and dark energy. Another component, presumably a scalar field, dominated during the period of inflation in the early Universe. In the following we shall first review the equilibrium thermodynamics of relativistic and non-relativistic particles and then proceed to the brief thermal history of the Universe. Let us consider a particle with momentum p and mass m. From Special Relativity the energy of this particle is E = p 2 + m2 , where p ≡ | p|. The phase space occupancy f ( p) in equilibrium at temperature T is given by the following distribution function 1 f ( p) = , (2.39) exp[(E − µ)/T ] ± 1 where µ is the chemical potential of each species. The plus and minus signs represent the Fermi–Dirac distribution and the Bose–Einstein distribution, respectively.
14
Expansion history of the Universe
Generally the distribution function f depends on the position x of the species, but Eq. (2.39) only depends on p ≡ | p| because of the homogeneity of the Universe. Since the minimum volume of phase space in terms of x and p is given by (2π )3 due to Heisenberg’s principle, the number of phase space elements is d3 x d3 p/(2π )3 . Then the energy density ρ and the pressure P with g∗ internal degrees of freedom are [73, 74] ∞ d3 p g∗ (E 2 − m2 )1/2 E(p) f (p) = dE ρ = g∗ E 2 , (2.40) (2π )3 2π 2 m exp[(E − µ)/T ] ± 1 d3 p pv d3 p p 2 f (p) = g f (p) P = g∗ ∗ (2π )3 3 (2π )3 3E ∞ (E 2 − m2 )3/2 g∗ dE = . (2.41) 6π 2 m exp[(E − µ)/T ] ± 1 Note that there is no integral over d3 x because ρ and P are defined as quantities per unit volume. In the first equality of Eq. (2.41) we have used the fact that the pressure per unit number density of particles is given by pv/3 (where v is the speed of the particle), and in the second equality of Eq. (2.41) the relation v = p/E has been used in unit of c = 1 (recall that the energyand the momentum 2 in Special Relativity are E = mc / 1 − v 2 /c2 and p = mv/ 1 − v 2 /c2 , respectively). In the final expressions of Eqs. (2.40) and (2.41) we have adopted the unit = 1. In what follows we consider relativistic and non-relativistic particles separately. (i) Relativistic species The relativistic limit corresponds to T m, i.e. taking the limit m → 0 in Eqs. (2.40) and (2.41). For non-degenerate particles (T µ) we obtain 2 (Bosons) (π /30)g∗ T 4 , (2.42) ρ= 2 4 (7/8)(π /30)g∗ T , (Fermions) P = ρ/3 , (2.43) ∞ 3 3 x ∞ where we have used 0 d x x /(e − 1) = π 4 /15 and 0 d3 x x 3 /(ex + 1) = 7π 4 /120. The result (2.43) shows that the equation of state of relativistic particles without degeneracies is given by w = 1/3. The photons are bosonic species and it is also known that the chemical potential µ for the CMB photons is much smaller than the temperature T (µ/T < 9 × 10−5 [75]). Since the photon has two spin states (g∗ = 2), its energy density is ργ =
π2 4 T . 15
(2.44)
2.3 Matter species in the Universe
15
The COBE satellite measured the present temperature of CMB photons to be T = 2.725 ± 0.002 K [12]. On using the conversion 1 K4 =1.279×10−35 g cm−3 , the energy density of CMB photons in the present Universe is ργ(0) = 4.641 × 10−34 g cm−3 . This corresponds to the density parameter (0) γ ≡
8π Gργ(0) 3H02
=
ργ(0) ρc(0)
= 2.469 × 10−5 h−2 .
(2.45)
−5 If we take the value h = 0.72, then (0) γ = 4.763 × 10 . Since the energy den−4 sity ργ evolves as ργ ∝ a (see Eq. (2.26) with w = 1/3), the comparison with Eq. (2.44) gives the relation T ∝ 1/a (= 1 + z). Hence the temperature is inversely proportional to the scale factor. Neutrinos also behave like relativistic particles provided that their masses are small. They are fermionic particles with zero chemical potentials and there are three types of species in standard models (electron neutrino νe , muon neutrino νµ , and tau neutrino ντ ). Each species has one spin degree of freedom. Note also that neutrinos have anti-particles (anti-neutrinos). Using Eq. (2.42) in the fermionic case, the energy density of neutrinos, including anti-particles, is given by
ρν = Neff
7π 2 4 T , 120 ν
(2.46)
where Neff is the effective number of neutrino species and Tν is the temperature of neutrinos. Note that Neff = 3 for standard models of neutrinos, but in the fermionic case we have introduced the effective number Neff in order to allow for other relativistic degrees of freedom. The Big Bang Nucleosynthesis (BBN) occurred around the energy scale ∼0.1 MeV to form light elements such as deuterium and helium. The decoupling of neutrinos from the rest of the cosmic plasma, immediately followed by the annihilation of electrons (e− ) and positrons (e+ ), occurred earlier than the BBN epoch. The presence of extra relativistic degrees of freedom changes the amount of the light elements predicted by the BBN, which allows to put a bound on Neff . The current standard value is Neff = 3.04 [76], which is slightly larger than 3. The neutrino temperature Tν is linked to the photon temperature Tγ via the relation Tν /Tγ = (4/11)1/3 . This comes from the conservation of entropy before and after the annihilation of electrons and positrons (see the problem 2.2). From Eqs. (2.44) and (2.46), the relation between the neutrino density and the photon density is ρν = Neff (7/8)(4/11)4/3 ργ . Hence the present density parameter of radiation, which is the sum of photons and relativistic neutrinos, yields (0) r =
ργ(0) + ρν(0) ρc(0)
= (0) γ (1 + 0.2271Neff ) ,
(2.47)
16
Expansion history of the Universe
where (0) γ is given in Eq. (2.45). If we take the values h = 0.72 and Neff = 3.04, −5 then we obtain (0) r = 8.051 × 10 . (ii) Non-relativistic matter In the case of non-relativistic particles (T m), Eqs. (2.40) and (2.41) reduce to mT 3/2 exp [−(m − µ)/T ] , (2.48) ρ = g∗ m 2π mT 3/2 T P = g∗ T (2.49) exp [−(m − µ)/T ] = ρ , 2π m which are valid for both bosonic and fermionic particles. See problem 2.3 for the derivation of (2.48) and (2.49). Equation (2.49) shows that the pressure P is suppressed relative to the energy density ρ by the factor T /m 1. Hence the equation of state for non-relativistic matter is w 0, as expected. The above result shows that the energy density ρ is not described by a function of the temperature T only (unlike the case of photons). Hence we need to measure the density of non-relativistic particles (baryons and dark matter) directly from observations. Let us consider baryons first. During the BBN epoch the light elements such as deuterium and helium were formed from neutrons and protons. Most of the neutrons decayed to protons (through the β-decay) before the formation of deuterium, while neutrons that did not decay to protons were eventually trapped in helium. If we increase the baryon density, the process of the BBN occurs faster and hence more neutrons remain without decaying to protons. This leads to an increase of the abundance of helium, whereas the abundance of deuterium decreases. Thus the amount of light elements produced during the BBN epoch is sensitive to the baryon density. The abundance of deuterium is known by observing absorption lines in the high-redshift quasars. According to the measurement of distant quasars, Tytler and his collaborators derived the primeval deuterium abundance relative to the hydrogen to be D/H = (3.0 ± 0.4) × 10−5 [77]. From this bound, Burles, Nollett, and Turner [78] obtained the following constraint on the present density parameter of baryons: 2 (0) b h = 0.020 ± 0.002
(BBN constraint),
(2.50)
at the 95% confidence level. The CMB observations also place tight bounds on the density parameter (0) b . If we increase the baryon density ρb , this leads to a smaller sound speed cs for the combined fluid system of baryons, photons, and electrons [see Eqs. (4.180) and (4.173)]. Crudely speaking the perturbations in CMB anisotropies with comoving 2 wavenumber k satisfy the equation for the harmonic oscillator, ddηx2 + k 2 cs2 x = 0, with some corrections [see Eq. (5.16)]. For smaller cs , the frequency kcs decreases
2.3 Matter species in the Universe
17
so that the height of the first CMB acoustic peak gets larger (because the amplitude of the harmonic oscillator with a smaller spring constant gets larger). One can constrain the amount of the baryon density by using this property. From the WMAP 5-year data combined with SN Ia and BAO data, the constraint on the present density parameter of baryons is [15] +0.00058 2 (0) b h = 0.02267−0.00059
(WMAP 5 year constraint),
(2.51)
at the 68% confidence level. If we take the value h = 0.72, then we have (0) b = 0.0437 for the central value in Eq. (2.51). This means that the baryonic contribution is only 4% in the present Universe. In addition to baryons, astrophysical observations require the existence of dark matter as another non-relativistic component in the Universe. Since dark matter interacts very weakly with standard model particles, its existence can only be probed by gravitational effects on visible matter. More specifically, if dark matter was non-relativistic at the time it decoupled from photons, it is called Cold Dark Matter (CDM). Alternatively, dark matter that was relativistic at the photon decoupling epoch is called Hot Dark Matter (HDM), whose representative candidate is the neutrino. The present paradigm of structure formation is based on the gravitational clustering of CDM. The baryonic matter alone is not sufficient to lead to structure formation consistent with observations of galaxy clustering [74]. Also the pure HDM model is ruled out as a viable model. This comes from the fact that neutrinos tend to stream out of any overdense region so that the CMB spectrum in the neutrino-dominated Universe has an insufficient power on small scales to be consistent with observations. In the mixed dark matter models of CDM and HDM, the observations limit the amount of hot dark matter to at most a few percent. In the following we shall focus on the pure CDM model (plus dark energy) unless otherwise stated. The CMB anisotropy data show that the present abundance of dark matter is about 5 times larger than that of baryons. The WMAP 5-year data constrain the density parameter of the CDM to be [15] 2 (0) c h = 0.1131 ± 0.00034
(WMAP 5-year constraint),
(2.52)
at the 68% confidence level. For the value h = 0.72 we have (0) c = 0.2182 for the central value in Eq. (2.52). The origin of dark matter has not been identified yet. There are basically two classes of dark matter – (i) the astrophysical candidates, or (ii) the particle candidates. Examples of class (i) are black holes, neutron stars, and white dwarfs. However, since these originate from baryons, it is not possible to explain all dark matter components without taking into account non-baryonic dark matter.
18
Expansion history of the Universe
Class (ii) is of the non-baryonic type. Some examples of this class are axions and Weakly Interacting Massive Particles (WIMPs, including neutralinos). The axion was originally introduced by Peccei and Quinn [79] as a solution to the strong CP problem in quantum chromodynamics (QCD). It has a weak coupling with a small mass ma = 10−6 –10−2 eV. The initial momentum of the axion when it gains a mass through non-perturbative QCD effects is of the order of pa = 10−9 eV ma . Hence the axion can be a good candidate for CDM. The WIMPs are usually motivated by supersymmetric theories. For example, neutralinos are formed as four eigenstates of a mass operator as a result of the mixing of superpartners of Z-bosons, the photon, and the neutral higgs (zino, photino, and higgsino, respectively). The lightest of the four neutralinos turns out to be the lightest supersymmetric particles with typical masses mn = 100 GeV–1 TeV. The lightest neutralinos couple to other particles with the strength characteristic of the weak interaction and hence they can be a good candidate for CDM. Direct or indirect dark matter searches and future LHC experiments will hopefully detect dark matter from space or from high-energy collisions of particles (see Refs. [19, 20, 21] for recent reviews). (iii) Dark energy From Eqs. (2.47), (2.51), and (2.52) the sum of the density parameters of radiation, baryons, and dark matter does not exceed 0.3 in the present Universe. Since present observational bounds on the spatial curvature are very strong, |(0) K | 0.01, we still need to identify the remaining 70% of the cosmic matter. This unknown component, called dark energy, is supposed to be responsible for the present cosmic acceleration. The combined data analysis using WMAP, SN Ia, and BAO have provided the following constraint for the present density parameter of dark energy: (0) DE = 0.726 ± 0.015
(WMAP 5-year constraint).
(2.53)
In Chapters 5 and 14 we will discuss observational constraints on dark energy in great detail. The theoretical attempts to identify the origin of dark energy will be discussed in Chapters 6–10. 2.4 Cosmic distances In order to discuss observational constraints on dark energy, it is important to introduce cosmic distances directly related to observations in the FLRW spacetime (2.1). In fact, a large part of the evidence for dark energy comes from measurements of cosmological distances. Setting r = sin χ (K = +1), r = χ (K = 0), and r = sinhχ (K = −1) in Eq. (2.2), the 3-dimensional space line-element is expressed as dσ 2 = dχ 2 + (fK (χ ))2 dθ 2 + sin2 θ dφ 2 , (2.54)
2.4 Cosmic distances
where
sinχ fK (χ ) = χ sinhχ
(K = +1) , (K = 0) , (K = −1) .
The function (2.55) can be written in a unified way: √
1 fK (χ ) = √ sinh −Kχ , −K
19
(2.55)
(2.56)
where the case of the flat universe is recovered by taking the limit K → −0. 2.4.1 Comoving distance Let us first compute the comoving distance dc . The light traveling along the χ direction satisfies the geodesic equation: ds 2 = −c2 dt 2 + a 2 (t)dχ 2 = 0, where we have recovered the speed of light c for clarification. Let us consider the case in which light emitted at time t = t1 with χ = χ1 (redshift z) reaches an observer at time t = t0 with χ = 0 (corresponding to z = 0). Integrating the equation, dχ = −c dt/a(t), the comoving distance reads χ1 t1 c dt . (2.57) dχ = − dc ≡ χ1 = 0 t0 a(t) From Eq. (2.28) it follows that dt = −dz/[H (1 + z)]. Then the comoving distance is given by z c d˜z dc = , (2.58) a0 H0 0 E(˜z) where
The function, z 0
z 0
E(z) ≡ H (z)/H0 .
(2.59)
d˜z/E(˜z), can be expanded around z = 0:
d˜z E (0) 2 1 2 =z− z + 2E (0) − E (0) z3 + O(z4 ) , E(˜z) 2 6
(2.60)
where a prime represents a derivative with respect to z. For a redshift z much smaller than unity, the comoving distance is approximately given by c z , for z 1 . (2.61) dc a0 H 0 On using the relation (2.29), we find v (a0 H0 )dc .
(2.62)
20
Expansion history of the Universe
This shows that the recessional velocity v of the object is proportional to dc with the proportionality constant a0 H0 . For the physical distance r = a0 dc we find r (c/H0 )z v/H0 , which means that Hubble’s law (2.32) is satisfied. Hubble’s law written as in Eq. (2.32) is valid therefore only in the low-redshift region z 1. For z 1 the higher-order terms in Eq. (2.60) become important so that Hubble’s law is subject to be modified. 2.4.2 Luminosity distance As we will see in Section 5.2, the luminosity distance dL is used in SN Ia observations in order to link the supernova luminosity with the expansion rate of the Universe. It is defined by Ls , (2.63) 4π F where Ls is the absolute luminosity of a source and F is an observed flux. Note that the observed luminosity L0 (detected at χ = 0 and z = 0) is different from the absolute luminosity Ls of the source (emitted at the comoving distance χ with the redshift z). The flux F is defined by F = L0 /S, where S = 4π (a0 fK (χ ))2 is the area of a sphere at z = 0. Then the luminosity distance (2.63) yields dL2 ≡
dL2 = (a0 fK (χ ))2
Ls . L0
(2.64)
We need now to derive the ratio Ls /L0 . If we write the energy of light emitted at the time-interval t1 to be E1 , the absolute luminosity is defined by Ls = E1 /t1 . Similarly the observed luminosity is given by L0 = E0 /t0 , where E0 is the energy of light detected at the time-interval t0 . Since the energy of a photon is inversely proportional to its wavelength λ we have that E1 /E0 = λ0 /λ1 = 1 + z, where we have used Eq. (2.28). Moreover, the constancy of c = λ/t implies λ1 /t1 = λ0 /t0 , where λ1 and λ0 are the wavelength of light at the points of emission and detection respectively. This leads to the relation t0 /t1 = λ0 /λ1 = 1 + z. Hence we find Ls E1 t0 = = (1 + z)2 . L0 E0 t1
(2.65)
From Eqs. (2.64) and (2.65) the luminosity distance reduces to dL = a0 fK (χ)(1 + z) .
(2.66)
Recall that the function fK (χ ) is given in Eq. (2.56) with the comoving distance z d˜z c χ = dc = . (2.67) a0 H0 0 E(˜z)
2.4 Cosmic distances
Then dL can be expressed as c(1 + z) sinh dL = H0 (0) K
(0) K
21
z 0
d˜z E(˜z)
,
(2.68)
(0) 2 2 where (0) K = −Kc /(a0 H0 ) . Note that this definition of K is identical to the last one given in Eq. (2.24) in the unit c = 1. It is clear that the luminosity distance is directly related to the expansion rate of the Universe. Expanding the function sinh(x) in the form sinh(x) = x + x 3 /6 + O(x 5 ) and using Eq. (2.60), we find that dL can be expanded around z = 0 as follows: E (0) 2 c dL = z+ 1− z H0 2 1 2 (0) 3 4 + 2E (0) − 3E (0) − E (0) + K z + O(z ) . (2.69) 6
In the small-redshift region (z 1) we have dL c z/H0 . Using Eq. (2.29) we obtain v H 0 dL ,
for
z 1.
(2.70)
This shows that Hubble’s law holds for the luminosity distance as well. 2.4.3 Angular diameter distance The angular diameter distance dA is defined by x , (2.71) θ where θ is the angle that subtends an object of actual size x orthogonal to the line of sight. This distance is often used for the observations of CMB anisotropies. Since the source lies on the surface of a sphere with radius χ with the observer at the center, the size x at time t1 in the FLRW spacetime (2.1) with (2.54) is given by dA ≡
x = a(t1 )fK (χ )θ . Hence the diameter distance is
(2.72)
z c d˜z (0) , (2.73) sinh K z) 0 E(˜ H0 (0) K z where we have used fK = c/(a0 H0 (0) ) sinh ( (0) z/E(˜z)) and z = K K 0 d˜ a0 /a(t1 ) − 1. Comparing Eq. (2.73) with Eq. (2.68), we notice the 1 a0 fK (χ ) = dA = a(t1 )fK (χ) = 1+z 1+z
22
Expansion history of the Universe
following relation dA =
dL . (1 + z)2
(2.74)
This is called reciprocity or duality or Etherington relation [80]. Its validity extends far beyond the FLRW metric: it is valid in fact for any metric as long as flux is conserved. In the limit z 1 all the distances discussed above reduce to the Euclidean distance in the Minkowski spacetime. 2.4.4 Degeneracy of the distance–redshift relation All the distance definitions given above z depend on the cosmological parameters through the integral χ = (c/(a0 H0 )) 0 d˜z/E(˜z) in flat spaces and on fK (χ ) in curved spaces. It is therefore clear that all the measures of expansion that we can obtain through measurements of distances, from standard candles to the CMB acoustic peaks, will constrain only the cosmological parameters contained in E(z) and only in those particular combinations that appear in χ and in fK (χ ). If we had distance information only for a given z then all the combinations of cosmological parameters that produce the same fK (χ ) would be equally acceptable: the constraints would therefore be fully degenerate along lines (or surfaces) of constant fK (χ ). For instance the closed CDM cosmological model defined by (0) (0) (0) ((0) , m ) = (1, 1) and the open model defined by ( , m ) = (0.1, 0.6) give practically identical distances at z = 1. In Fig. 2.1 we plot the lines of constant (0) fK (χ) in the plane ((0) , m ) for redshifts that roughly correspond to a typical distant supernova and to CMB. If we have information only in a small range of redshifts the degeneracy will be partially broken but still the constraints will appear elongated along the lines of equal distances, as we will see in the next sections. It is only by combining measures at widely different redshifts or by employing indicators other than distances that we may hope to pin down the cosmological parameters. 2.5 The equation of state of dark energy Let us consider the Universe filled by radiation (density ρr and pressure Pr = ρr /3), non-relativistic matter (density ρm and pressure Pm = 0), and dark energy (density ρDE and pressure PDE ). Since ρr and ρm evolve as ρr ∝ a −4 and ρm ∝ a −3 , respectively, they can be expressed in the forms ρr = ρr(0) (a0 /a)4 = ρr(0) (1 + z)4 ,
(2.75)
ρm = ρm(0) (a0 /a)3 = ρm(0) (1 + z)3 .
(2.76)
2.5 The equation of state of dark energy 3
3
1
0
1 0.0
bi g
ba ng
1
no
ba ng
no
2
0
0
z 1100
bi g
z 1 2
23
0
0.5
1.0
1.5
2.0
2.5
1 0.0
3.0
0.5
0 m
1.0
1.5
2.0
2.5
3.0
0 m
Figure 2.1 Contours of constant fK (χ ) for z = 1 and z = 1100. The straight line represents the flat model. In the left panel the contours range from 1400 to 2800 h−1 Mpc in steps of 200, from right to left. In the right panel they range from 300 to 5100 h−1 Mpc in steps of 600, from top to bottom (the last two contours are for 15 000 and 30 000 h−1 Mpc, respectively).
The redshift zeq that corresponds to the radiation–matter equality (ρr = ρm ) is 1 + zeq =
ρm(0) ρr(0)
=
(0) m (0) r
,
(2.77)
(0) where (0) r is given by Eq. (2.47) with Eq. (2.45). The density parameter m (0) is the sum of the baryon contribution b and the CDM contribution (0) c , i.e. (0) (0) (0) (0) (0) = + . The WMAP 5-year constraints on and are given by m c c b b Eqs. (2.51) and (2.52), respectively. For the effective number of neutrino species Neff = 3.04 we obtain 2 1 + zeq = 2.396 × 104 (0) m h .
(2.78)
2 If we take the value (0) m h = 0.136, we have zeq = 3258. Note that the CMB decoupling epoch corresponds to zdec 1090 [15], therefore later than the radiation– matter equality. Let us consider dark energy with an equation of state wDE = PDE /ρDE , satisfying the continuity equation
ρ˙DE + 3H (ρDE + PDE ) = 0 .
(2.79)
Integrating this equation by using the relation dt = −dz/[H (1 + z)], we obtain z 3(1 + wDE ) (0) ρDE = ρDE exp d˜z , (2.80) 1 + z˜ 0
24
Expansion history of the Universe
which can also be written by introducing an average wˆ DE as z 1 wDE (˜z) (0) −3(1+wˆ DE ) d˜z . (2.81) , wˆ DE (z) = ρDE = ρDE (a/a0 ) ln(1 + z) 0 1 + z˜ From the Friedmann equation (2.17) we have H2 =
K 8π G (ρr + ρm + ρDE ) − 2 . 3 a
(2.82)
From Eq. (2.82) we see that the present density parameters defined in Eq. (2.24) obey the following relation (0) (0) (0) (0) r + m + DE + K = 1 .
(2.83)
Then Eq. (2.82) can be written in the form 2 2 4 (0) 3 H (z) = H0 (0) r (1 + z) + m (1 + z) + (0) DE
exp 0
z
3(1 + wDE ) (0) 2 d˜z + K (1 + z) . 1 + z˜
(2.84)
Differentiating this equation with respect to z, we find that the equation of state of dark energy can be expressed as (0) 4 2 (1 + z)(E 2 (z)) − 3E 2 (z) − (0) r (1 + z) + K (1 + z) , wDE (z) = (0) (0) 4 3 2 3 E 2 (z) − (0) r (1 + z) − m (1 + z) − K (1 + z)
(2.85)
where E(z) is defined in Eq. (2.59) and a prime represents a derivative with respect to z. From Eq. (2.68) the quantity E(z) can be written in terms of dL : 2 2 H d (z) (1 + z)2 c2 (1 + z)2 + (0) L K 0 . (2.86) E 2 (z) = [(1 + z)H0 dL (z) − H0 dL (z)]2 For the flat Universe ((0) K = 0) this relation reduces to the following simple form d dL (z) −1 c E(z) = . H0 dz 1 + z
(2.87)
If the luminosity distance dL (z) is measured observationally, we can determine the evolution of E(z) from Eq. (2.86) and hence wDE (z) from Eq. (2.85). The cosmic expansion history for the redshift z O(1) can be reconstructed from the SN Ia observations. In this regime the energy density of radiation is negligible compared to those of non-relativistic matter and dark energy. The present
2.5 The equation of state of dark energy
25
observational bound on the cosmic curvature is −0.0175 < (0) K < 0.0085 [15], showing that the Universe is close to the flat geometry. In the flat Universe with a negligible contribution of radiation, Eq. (2.85) reduces to wDE (z) =
(1 + z)(E 2 (z)) − 3E 2 (z) . 3 3 E 2 (z) − (0) (1 + z) m
(2.88)
This relation is often used when we place observational constraints on the equation of state of dark energy. While the effect of the cosmic curvature on the estimate of wDE (z) can be negligible in the region z 1, a small uncertainty on the curvature can produce a significant bias in wDE in the high-redshift regime z 1 [81, 82]. In such a case we need to use the relation (2.85) rather than (2.88). It is important to remark however that the basic observable quantity is E(z), not wDE (z). In fact, wDE (z) cannot be determined entirely from E(z), i.e. from measurements of the background expansion. From Eq. (2.88) it appears in fact that one needs (0) m , i.e. the present density of pressureless matter, and this can only be obtained from large-scale structure methods. However one must notice that the density parameter obtained from e.g., the cluster mass estimation does not necessarily coincide with the quantity (0) m in Eq. (2.88), since in general clustered matter and pressureless matter do not need to be the same. This is particularly important to notice in coupled dark energy models [17] in which matter acquires an effective pressure through an interaction with dark energy. Of course if wDE (z) is assumed to be constant or is parametrized in some form, as is usually done, then the knowledge of E(z) at several z’s can fix both the equation of state and (0) m . Finally, let us notice that the bound on the equation state of dark energy from the WMAP 5-year data combined with other observational data is −1.097 < wDE < −0.858 at the 95% confidence level [15]. Hence we cannot rule out the possibility that wDE is smaller than −1. These cases are generally called “phantoms” or “ghosts” [83]. Since P + ρ < 0 in this case, Eq. (2.20) shows that ρ increases with time. From Eq. (2.17) the Hubble parameter H grows toward the future. When wDE is constant and smaller than −1, the solution of the scale factor corresponding to the expanding Universe is given by a ∝ (trip − t)2/(3(1+wDE )) ,
(2.89)
where t is smaller than the constant trip . As t approaches trip , the scale factor goes to infinity. One can easily show that the scalar curvature R and the Hubble parameter H also diverge at t = trip . The Universe ends up with a finite-time singularity in the (distant!) future (see problem 2.4). This finite-time singularity is called the big-rip singularity [84].
26
Expansion history of the Universe
2.6 Problems 2.1 Derive the continuity equation, ρ˙ + 3H (ρ + P ) = 0, from the continuity equation ∇µ T0µ = 0 .
(2.90)
2.2 Let us consider the entropy conservation before and after the annihilation of electrons and positrons when the cosmic temperature was of the order of the electron mass. The entropy density for a particle with density ρ, pressure P , and temperature T is defined by s ≡ (ρ + P )/T . Before the annihilation there were photons, neutrinos, anti-neutrinos, electrons, and positrons with the same temperature, whereas after the annihilation there were photons, neutrinos, and anti-neutrinos with different temperatures. By using the entropy conservation as well as the fact that the neutrino temperature scales as Tν ∝ a −1 , show that the relation between the photon temperature Tγ and the neutrino temperature Tν is given by 1/3 4 Tν = . (2.91) Tγ 11 2.3 Derive the density (2.48) and the pressure (2.49) in the non-relativistic limit (T m). 2.4 Show that the scalar curvature R diverges at the big-rip singularity by using the solution (2.89). For the dark energy equation of state wDE = −1.5, estimate the time trip − t0 by neglecting the contribution of non-relativistic matter.
3 Correlation function and power spectrum
The models of dark energy we will introduce later on are linked to the observations by a precious tool, statistics. Since this world is complicated, we have to average the ups and downs of everyday life to get a sense of the underlying substance. In this chapter we present basic tools of statistics in order to confront dark energy models with observations. Statistics itself is often divided into descriptive statistics, i.e. how to condense the data in a compact and useful way, and estimation (or inferential, inductive) statistics, i.e. how to derive information on model parameters. We start here with the statistics needed for cosmological perturbation theory, essentially descriptive statistics such as correlation function and power spectrum, and postpone parameter estimation statistics to Chapter 13. A note on notation is in order here. When there is no real need, we will not use separate notation for an estimator (say, the correlation function) and its expected value. Similarly, when there is no risk of confusion, we will denote vector quantities like position x and wavevector k with unbolded fonts x, k, especially to denote the argument of functions: δ(x) will in general mean the density contrast at a position x. Finally, when an integration, even a multiple one, is extended to the whole space we simply write dV or dx. When the domain of integration really matters, then it will be specified. 3.1 The correlation function Our first task is to describe a random distribution of particles in a compact way, to be identified with astrophysical sources (galaxies, quasars, etc.). If there are N points in a volume V , the first interesting descriptor is the average numerical density ρ0 = N/V . Clearly this is utterly insufficient to discriminate among say N points clustered near the same spot and N points evenly distributed across the volume, so we need to find more useful descriptors. Let us focus then on some 27
28
Correlation function and power spectrum
small volume dV chosen randomly inside the volume V . Then ρ0 dV is the average number of particles in the infinitesimal volume. If dNab = na nb is the average number of pairs in the volumes dVa and dVb (i.e. the product of the number of particles in one volume times the number in another volume), separated by rab , then we define the next important descriptor, the 2-point correlation function ξ (rab ), as dNab = na nb = ρ02 dVa dVb [1 + ξ (rab )] .
(3.1)
We have implicitly assumed that rab > 0, i.e. the two volumes do not coincide. The word “average” (and the symbol we will use to denote it) may have two meanings. One can average by taking many realizations of the distribution, all of them produced in the same way (e.g., by an N -body computer code or by throwing particles at random), selecting in each realization the volumes dVa , dVb at the same locations and then averaging the pair number na nb . This is the ensemble average. Alternatively one can take the pairs at different spots within the same realization, separated by the same rab (sample average). If the spots are so distant that they are uncorrelated, then we can consider them as coming from different realizations and the two averaging methods coincide. The problem with the approach is that we do not know a priori whether the spots really are uncorrelated until we can compare them with an ensemble of realizations. This issue is particularly acute in astrophysics since we are given a single Universe. The sample correlation function does not in general coincide with the one we would obtain from an ensemble. This problem is of course more important for distributions that are inhomogeneous at very large scales. The estimation of the correlation function on scales smaller than the scale of (approximate) homogeneity will not coincide with the ensemble value. See problem 3.1 for an example. Even in those cases, however, the correlation function is a useful descriptor (although a survey-dependent one) and it makes sense to use it. However here we will always assume that the properties of the sample distribution are a good approximation of the ensemble ones. If the distribution has been obtained by throwing the N particles at random (i.e. without any preference with respect to the place), then there is no reason for dNab to depend on the location. Therefore the average number of pairs is exactly equal to the product of the average number of particles in the two volumes, na nb = na nb = ρ02 dVa dVb , and the correlation ξ vanishes. Conversely, if ξ is non-zero, we say that the particles are correlated. Then the correlation function can be written as a spatial average of the product of the density contrast δ(ra ) = na /(ρ0 dVa ) − 1 at two different points: ξ (rab ) =
dNab 2 ρ0 dVa dVb
− 1 = δ(ra )δ(rb ) ,
(3.2)
3.1 The correlation function
29
where we have used δ(ra ) = δ(rb ) = 0. If this average is taken to be the sample average, then it means we have to average over all possible positions: 1 (3.3) ξ (r) = δ( y)δ( y + r) dVy . V A bit of jargon: when ξ (r) depends only on the separation r and not on r a and r b individually, the system is said to be statistically homogeneous (i.e. it possesses the same statistical properties everywhere), although strictly speaking for this property to be true all the higher-order statistics should also be independent of location. If moreover the ensemble average coincides with the sample average, then the system is said to be ergodic. However the latter term refers historically to time processes, not to spatial ones. The term most often used in astrophysics is that the system is a fair sample of the Universe. In practice it is easier to derive the correlation function as the average density of particles at a distance r from another particle, i.e. by choosing the volume dVa so that ρ0 dVa = 1. Then the number of pairs is given by the number of particles in the volume dVb : dNb = ρ0 dVb [1 + ξ (rb )] .
(3.4)
Operationally therefore one evaluates the correlation function as follows: ξ (r) =
ρc dN (r) −1= − 1, ρ0 dV ρ0
(3.5)
i.e. as the average number of particles at distance r from any given particle (or number of neighbors), divided by the expected number of particles at the same distance in a uniform distribution, minus 1 (sometimes this is called conditional density contrast). In a finite volume with N = ρ0 V particles one clearly has an integral constraint on ξ (r) due to the fact that the average density is calculated within the volume itself: dN N 1 dV − V = − V = 0. (3.6) ξ (r)dV = ρ0 dV ρ0 If the correlation ξ (r) is positive, there are more particles than in a uniform distribution. In this case the distribution is said to be positively clustered. Quite often one is interested only in the dependence on the modulus r, so the volume at distance r is chosen as a shell around each particle. One could generalize this definition by introducing the anisotropic correlation function as the number of pairs in volumes separated by the vector r. This is useful whenever there is some reason to suspect that the distribution is indeed anisotropic, as when there is a significant distortion along the line of sight due to the redshift.
30
Correlation function and power spectrum
The estimator (3.5) requires the knowledge of the number density ρc inside a shell of thickness dr at distance r from every particle. In other words, it requires the estimation of the density in every shell. In practice, a direct estimation of the shell density is difficult because of the complicated boundary and selection procedure that a real survey often has. The simplest way to measure ξ is to compare the real catalog to an artificial random catalog with exactly the same boundaries and the same selection function. Then the estimator can be written as ξ=
DD − 1, DR
(3.7)
where DD means the number of galaxies at distance r counted by an observer centered on a real galaxy (data D). This is divided by the number of galaxies DR at the same distance but in the random catalog (if to reduce the scatter the random catalog contains α times more galaxies than the real one then DR has to be divided by α). In other words, instead of calculating the volume of the shell (which is a difficult task in realistic cases), we estimate ξ by counting the galaxies in the Monte Carlo realization. In this way all possible boundaries and selection function can be easily mimicked in the random catalog, which will affect DD and DR in the same way. Following these considerations, similar estimators with numerically more robust properties have been proposed [85].
3.2 The n-point correlation function The average number density ρ0 can be called a one-point estimator of a random field, since it is estimated by averaging an ensemble over the same location. The correlation function we have just seen is a two-point estimator because it requires averaging over two small volumes. It is clear that one can extend these definitions to higher orders. For instance, the three-point function is defined as ςabc (ra , rb , rc ) = δ(ra )δ(rb )δ(rc ) .
(3.8)
In terms of the counts in infinitesimal cells, we can write na nb nc ςabc (ra , rb , rc ) = −1 −1 −1 ρ0 dVa ρ0 dVb ρ0 dVc =
na nb nc 3 ρ0 dVa dVb dVc
− ξab − ξbc − ξac − 1 ,
(3.9)
where ξij ≡ ξ (rij ). We then obtain the following useful relation na nb nc = ρ03 dVa dVb dVc (1 + ξab + ξbc + ξac + ςabc ) ,
(3.10)
3.3 The power spectrum
31
where ςabc is called the “disconnected” part of the third-order correlation function. A random field is said to be Gaussian when ςabc = 0 (and the same for all disconnected higher-order moments). In this case the two-point correlation function (or its Fourier version, the power spectrum, see the next section) completely describes the statistical properties of the field.
3.3 The power spectrum As we will see in the next chapters, in particular in Chapter 4, the linear perturbation variables contain important physics, both for dark energy and cosmology in general. A convenient way to study perturbation variables is to decompose fluctuations into orthonormal modes because at the linear level they evolve independently. Since by definition the average of a perturbation variable is zero, the simplest non-trivial statistics corresponds to a quadratic function of the variables. In Fourier space, any real quadratic function of a perturbation variable is called a power spectrum. Examples are Pδ (k) = A|δk |2 ,
(3.11)
P (k) = B|k |2 ,
(3.12)
where δk and k are the Fourier coefficients of the density contrast and the gravitational potential, respectively, with A and B being some constants. If the quadratic form is composed of two different variables, e.g., |δk θk |, then we have a crosscorrelation power spectrum. If we average over directions, the power spectrum will depend only on the modulus k. The power spectrum is by far the most common descriptor of clustering in the linear and mildly non-linear regime and plays a central role in cosmology and in this book. As we will show soon, the power spectrum is the Fourier-space version of the correlation spectrum. Unless otherwise specified, the normalization convention for the 3-dimensional (3D) Fourier transformation is V f (x) = (3.13) fk eik·x d3 k , (2π )3 1 f (x)e−ik·x d3 x . (3.14) fk = V With these conventions, f (x) and fk have the same dimensions. Remember however that for the normalization of the Fourier transformation the only important property is that the product of the pre-integral factors in 3 dimensions amounts to (2π )−3 . For all theoretical manipulations the factor of V can well be replaced by unity and often for convenience we will do so. Dirac’s delta function δD (x) is
32
Correlation function and power spectrum
defined as −3
δD (x) = (2π )
eik·x d3 k .
(3.15)
Analogous definition holds for Dirac’s function in Fourier space (which is not the Fourier transform of δD (x)) −3 (3.16) eik·x d3 x , δD (k) = (2π ) and their normalization is such that 3 δD (k)d k = δD (x)d3 x = 1 .
(3.17)
For the density contrast of a density field δ(x), the Fourier transform is 1 (3.18) δk = δ(x)e−ik·x dV . V The power spectrum is defined as P (k) = V |δk |2 = V δk δk∗ .
(3.19)
Notice that the power spectrum has the dimension of a volume. It follows that 1 P (k) = δ(x)δ( y)e−ik·(x− y) dVx dVy . (3.20) V Setting r = x − y, the spectrum (3.20) reduces to P (k) = ξ (r)e−ik·r dV , where ξ (r) is defined in Eq. (3.3): 1 ξ (r) = δ( y + r)δ( y) = V
(3.21)
δ( y + r)δ( y)dVy .
(3.22)
Therefore, the power spectrum is the Fourier transform of the correlation function (Wiener–Khinchin theorem). The converse property is −3 (3.23) ξ (r) = (2π ) P (k)eik·r d3 k . Notice that here and in the following the Fourier volume factor is not included, as in most literature. Assuming spatial isotropy, i.e. that the correlation function depends only on the modulus r = |r|, the spectrum depends only on k = |k|: π 2π sin kr 2 2 −ikr cos θ P (k) = ξ (r) r dr e sin θ dθ dφ = 4π ξ (r) r dr . kr 0 0 (3.24)
3.3 The power spectrum
33
The integral constraint (3.6) implies P (0) = 0 in k-space: in any finite survey, the spectrum vanishes at large scales. A more general definition of the power spectrum can also be given, but this time we have to think in terms of ensemble averages rather than volume averages. Consider in fact the ensemble average of V δk δk∗ : 1 V δk δk∗ = δ(x)δ( y)e−ik·x+ik · y dVx dVy V 1 = (3.25) δ( y)δ( y + r)e−i(k−k )· y−ik·r dVr dVy . V In order to perform ensemble averages, one has to think of fixing a position and taking the average over the ensemble of realizations. Then the average can enter the integration and acts only over the random variables δ. Then we obtain 1 V δk δk∗ = ξ (r)e−i(k−k )· y−ik·r dVr dVy V 1 −i(k−k )· y e = dVy e−ik·r ξ (r)dVr V =
(2π )3 P (k)δD (k − k ) . V
(3.26)
The definition (3.26) shows that modes at different wavelengths are uncorrelated if the field is statistically homogeneous (that is, if ξ does not depend on the position y but only on the separation r). Since δ(x) is a real function, δ(k) = δ ∗ (−k) and therefore the form V δk δk = (2π )3 V −1 P (k)δD (k + k ) can also be employed. In our cosmological applications we often wish to study a continuous underlying field, for instance the gravitational potential or the dark matter density contrast δm . We think of the galaxies as a mere discrete tracer of this field, as sand particles trace the wind in a sandstorm. The galaxies form therefore a discrete sampling of the underlying field. The only way to get information on the underlying field is to study this discrete sampling. Although we may assume by and large that “more galaxies” mean also “larger dark matter density contrast,” i.e. that the average of the galaxy density contrast δg is everywhere equal to δm , we need to take into account possible departure from the ideal case. The simplest case is when the number of galaxies in a given location follows a Poissonian distribution with the expected mean value δm . This would be the case if the galaxy formation process were indeed a Poissonian process: the outcome of many independent factors each with the same, small probability to occur. We do not know yet if this is an accurate description of what happens but at least it is a reasonable starting hypothesis.
34
Correlation function and power spectrum
To investigate the discreteness, we assume as field a collection of N particles of dimensionless unitary masses at positions x i , in a volume V . In the following we will make use of the window function W (x), a function that expresses the way in which the particles are selected. A typical selection procedure is to take all particles within a given region, and no particles elsewhere. In this case, the function will be a constant inside the survey, and zero outside (top-hat window function). We will always consider such a kind of window function in the following, and normalize it such that W (x)dV = 1 . (3.27) With this normalization, W (x) = 1/V inside the survey. The density contrast field we have in a specific sample is therefore the field times the window function (times the sample volume V because of the way we have normalized W ): δs = δ(x) V W (x) .
(3.28)
Let us now express the field as a sum of Dirac functions ρ(x) = i δD (x − x i ) so that ρ(x) V − 1 V W (x) = wi δD (x − x i ) − V W (x) , (3.29) δs (x) = ρ0 N i where wi = V W (x i ) and as usual ρ0 = N/V . The Fourier transform is V 1 1 δk = wi δD (x − x i ) − V W (x) e−ik·x dV = wi e−ik·x i − Wk , V N i N i (3.30) where we have introduced the k-space window function (3.31) Wk = W (x)e−ik·x dV , normalized such that W0 = 1. The spherical top-hat function corresponds to W (x) = 1/V ,
inside a spherical volume V of radius R ,
W (x) = 0 ,
outside .
We then have
(3.32)
3(sin kR − kR cos kR) r sin kr dr = , k (kR)3 0 V (3.33) where in the second equality we have integrated out the angular part as we did to derive Eq. (3.24). Notice that the function declines rapidly as k → π/R. Now Wk = V −1
e−ik·x dV =
3 R3
R
3.3 The power spectrum
35
squaring and averaging δk in Eq. (3.30) by separating the i = j terms from the others, we have 2 (k) ≡ V δk δk∗ = P (k) + Pn ,
(3.34)
where the “true” spectrum P (k) and the pure noise spectrum are given, respectively, by V wi wj e−ik·(x i −x j ) − V Wk2 , (3.35) P (k) = 2 N i=j Pn =
V 2 V , wi = N2 i N
(3.36)
where the last equality holds if wi equals 0 or 1. In order to derive Eq. (3.34) we have used the relation δ−k = 0 1 −ik·x i = Wk . wi e (3.37) N i The noise spectrum, negligible only for large densities, ρ0 = N/V → ∞, is the power spectrum of a distribution with no intrinsic correlation, i.e. obtained by throwing the particles at random. More exactly, it is the power spectrum of a Poissonian distribution. Since the galaxy distributions are often sparse, the noise is not always negligible and has to be subtracted from the estimate if we want to detect the underlying correlation. Therefore the estimator of the “true” power spectrum P (k) can be taken as Pˆ (k) = 2 (k) − Pn .
(3.38)
As for the correlation function, the power spectrum does not characterize a distribution completely, unless we know the distribution has some specific property, e.g., Gaussian, or Poisson, etc. In particular, if we assume the fluctuations to be Gaussian, we can derive the variance of the power spectrum, defined as σP2 ≡ [Pˆ (k) − P (k)]2 = 4 (k) − 2 (k)2 ,
(3.39)
where P (k) ≡ Pˆ (k) = 2 (k) − Pn . To evaluate 4 (k) we proceed as follows. Neglecting the window term (i.e. in the limit of a large volume) the continuous fluctuation field in Fourier space can be written in a compact way as gi δk = e−ik·r i , (3.40) 1/2 V i where by assumption gi are Gaussian random variables (the volume factor is only for convenience). The power spectrum is 2 (k) = V δk δk∗ = i gi2 . We want to
36
Correlation function and power spectrum
evaluate 2 (k)2 (k ) = V 2 δk δk∗ δk δk∗ =
gi gj gm gn e−i(k·r i −k·r j +k ·r m −k ·r n ) .
ij mn
(3.41) Now, the oscillating terms in the sum are negligible except when i = j = m = n or when the indices are equal in pairs. Then we can write gi4 + gi2 gj2 1 + e−i(k+k )·(r i −r j ) + e−i(k−k )·(r i −r j ) . 2 (k)2 (k ) = i
i
j =i
(3.42) Since gi is a Gaussian variable, all odd moments are zero and even moments can be written in terms of the variance; in particular, we have gi4 = 3gi2 2 . This is exactly what is needed to supply the sum for j = i with the missing terms j = i . Moreover, the oscillating exponential terms average out except for k close to k (or equivalently −k ). Then for k ≈ k we can neglect the k + k term since its faster oscillations are averaged over and therefore suppressed relative to the k − k term. Then we have 2 (k)2 (k ) = (3.43) gi2 gj2 1 + e−i(k−k )·(r i −r j ) , i
j
which amounts to 2 (k)2 (k ) = 2 (k)2 (k ) + V 2 δk δk∗ 2 .
(3.44)
From Eqs. (3.39) and (3.34) it follows that σP2 (k) = V 2 δk δk∗ 2 = (P (k) + Pn )2 .
(3.45)
We obtain finally for the fractional variance of the power spectrum with top-hat filtering a very simple and useful result 2 σP2 (k) 1 , (3.46) = 1+ P 2 (k) nP (k) where n = N/V is the number density. This tells us that, if the fluctuations are Gaussian, the error of the root mean square (rms) on the power spectrum is of the order of the power spectrum itself (including the shot noise). In general we consider the shell-averaged spectrum, i.e. the spectrum for all modes whose wavenumber modulus k lies within the shell k of volume Vk : 1 P (k )d3 k . (3.47) P (k) = Vk k
3.4 From the power spectrum to the moments
37
If the survey has a volume Vs = L3 , the lowest wavenumber we can safely construct is kmin = 2π/L. Then the number of independent k-modes in a volume Vk is Nk =
Vk V k Vs = . 3 kmin (2π )3
(3.48)
Therefore the error on the shell-averaged spectrum P (k) is reduced by the factor 1/Nk and we obtain 1 2 σP2 (k) (2π )3 1+ . (3.49) P 2 (k) Vk Vs nP Another way of looking at this equation is to say that the effective k-volume 3 3 degrades due to the shot noise to kmin (1 + 1/nP )2 , so that there are resolution kmin effectively less independent k-volumes to average over. A more complete derivation including an arbitrary density field and selection function is given in Ref. [86]. The general formula is (2π )3 Vs d3 r n4 w 4 [1 + 1/(nP (k))]2 σP2 (k) = , (3.50) P 2 (k) Vk [ Vs d3 r n2 w 2 ]2 where n = n(r) is the average density and w = w(r) is an arbitrary weight that might be modulated to minimize the variance itself. If n and w are constants, this expression reduces to Eq. (3.49). Finally one should be aware of the problem hidden in the result above. In order to evaluate the variance σP , the spectrum itself should be known. In practice, this means that one has to guess a spectrum before its variance can be evaluated. More consistently, in the likelihood method procedure of Chapter 13, we will take into account the fact that the variance depends on the model itself. 3.4 From the power spectrum to the moments Since the power spectrum is often the basic outcome of structure formation theories, it is convenient to express all the other quantities in terms of it. Here we find the relation between the power spectrum and the moments of the counts in random cells. Consider a finite cell. Divide it into infinitesimal cells with counts ni either zero or unity, so that for any positive power m we have nm (3.51) i = N0 , i
where N0 = ρ0 V is the count average. We have by definition of ξ for i = j : ni nj = ρ02 (1 + ξij )dVi dVj .
(3.52)
38
Correlation function and power spectrum
30000
30000
P(k) [(h/Mpc)3]
P(k) [(h/Mpc)3]
10000 7000
4000
10000
1000
5000
0.02
0.05 0.1 k [h/Mpc]
0.3
0.03
0.05 0.07 k [h/Mpc]
0.09
0.19
Figure 3.1 (Left) The power spectrum inferred from the SDSS galaxy power spectrum (data points from [65]) is compared to the predicted power spectrum based on the range of parameters consistent with the WMAP parameters (continuous lines). The galaxy power spectrum is normalized by weak lensing measurements [87]. (Right) The predicted power spectrum is compared to the mass power spectrum inferred from the 2-degree Field (2dF) Galaxy Redshift Survey [88]. From Ref. [14].
The count in the cell is N = then given by
i
ni . The variance (the second-order moment) is
M2 = N0−2 (N)2 = N0−2 (N 2 − N02 ) ,
(3.53)
where N = N − N0 . The expectation value N 2 is 2 2 2 ni ni + ni nj = N0 +N0 Wi Wj (1 + ξij )dVi dVj , nj = N = i
i=j
(3.54) where Wi is the window function and ξij ≡ ξ (|r i − r j |). We define the following integral (by definition Wi dV = 1 for any window function) 2 σ = W1 W2 ξ12 dV1 dV2 . (3.55) Then Eq. (3.54) reduces to N 2 = N0 + N02 (1 + σ 2 ) .
Since ξ12
= (2π )−3 P (k)eik·(r 1 −r 2 ) d3 k from Eq. (3.23), we have σ 2 = (2π )−3 P (k)eik·(r 1 −r 2 ) W1 W2 d3 k d3 r1 d3 r2 ,
(3.56)
(3.57)
3.5 Problems
39
where the integral dV1 dV2 in Eq. (3.55) is replaced by d3 r1 d3 r2 . For spherical cells of radius R, integrating over the angles, this reduces to 1 2 σR = P (k)WR2 (k)k 2 dk , (3.58) 2π 2 where we have used WR2 (k) = eik·r 1 W1 d3 r1 e−ik·r 2 W2 d3 r2 . If the cells have a radius of 8 h−1 Mpc, it turns out that σR is close to unity. Conventionally the normalization of the power spectrum is therefore given by quoting σ8 . Substituting Eq. (3.56) into Eq. (3.53), we find that the second-order moment M2 has the following relation with the power spectrum amplitude σ 2 : M2 = N0−1 + σ 2 .
(3.59)
The first and the second terms correspond to the noise and the count variance in the continuous limit, respectively. For the third-order moment we proceed in a similar fashion and we obtain (see problem 3.3) M3 = N0−2 + Wi Wj Wk ςij k dVi dVj dVk . (3.60) Of course similar relations can be found at any order. Non-zero higher-order moments are useful to quantify the deviation from Gaussianity of the matter and galaxy distribution. 3.5 Problems 3.1 Find the 3-dimensional correlation function of points distributed randomly on the equatorial plane of a sphere. 3.2 Using Eq. (3.23), find the isotropic correlation function of a distribution whose isotropic power spectrum is Gaussian with k ≥ 0 peaked at k = 0 with variance σk2 . 3.3 Derive Eq. (3.60).
4 Basics of cosmological perturbation theory
In this chapter we present the basics of linear perturbations in cosmology. After a general introduction of cosmological perturbation theory, we work out several cases: (i) single pressureless perfect fluid, (ii) single general perfect fluid, and (iii) two fluids: matter and radiation. We also discuss a number of topics such as the velocity field, the redshift distribution, Boltzmann equations, the matter power spectrum, and the perturbed photon propagation. These provide us with important tools when we confront dark energy models with observations of the cosmic microwave background (CMB) and large-scale structure (LSS). In this chapter the treatment and notation are fairly standard and the topic is covered in most modern textbooks on cosmology. Readers familiar with cosmological perturbation theory may skip this chapter.
4.1 Perturbing General Relativity In Chapter 2 we have outlined the cosmic expansion history in the homogeneous and isotropic FLRW background. However, our Universe is far richer than this simple picture. A metric that deviates from the FLRW spacetime can be written as the sum of an unperturbed FLRW part plus something else, that we can generally call “perturbed” metric. If the perturbed part is assumed to be small, in a sense to be defined later, then this splitting of the full metric into a background part and a perturbed one leads to extremely useful results. As we all know, physics is to a large extent described by a Taylor expansion to some low order, and cosmology is not an exception. To perturb the relativistic equations one must first of all perturb the metric, writing at first order (0) + δgµν , gµν = gµν
40
(4.1)
4.1 Perturbing General Relativity
41
where all the entries in the perturbed metric δgµν have to be small with respect to the 0-th order part. In this chapter we write the metric directly in terms of the conformal time η = a −1 dt. We consider cosmological perturbations about the flat FLRW metric given by (0) dx µ dx ν = a 2 (−dη2 + δij dx i dx j ) . ds 2 = gµν
(4.2)
We will also use the conformal Hubble quantity H≡
1 da = Ha . a dη
(4.3)
In General Relativity the field equations are invariant under a general coordinate change. This means that the split between a background metric and a perturbed one is not unique. However, although it is often a great simplification to choose some special coordinate frame, it would be very confusing if we change in the process also the unperturbed (or background) metric. We would like for instance to keep the FLRW metric as “the” background whenever we make a general transformation. (0) as it is, Therefore we select a class of infinitesimal transformations that leaves gµν while the perturbed metric δgµν is subject to change. Following general physics usage, this class of transformations is called gauge transformations. In the unperturbed Universe, we have already defined comoving coordinates in such a way that the matter particles expanding with the Universe remain at fixed (comoving) coordinates. When perturbations are added, we can either use the same coordinates, or set up a new set of coordinates that free-fall with the particles in the perturbed gravitational field, or even adopt a totally different frame not related to matter particles. That is, for instance, we can choose to attach the observers to the points in the unperturbed frame or to the perturbed particles. In the former case, to be called the Newtonian or longitudinal gauge, the observers will detect a velocity field of particles falling into the clumps of matter and will measure a gravitational potential. This choice is in fact the most intuitive one and reduces easily to the Newtonian case. On the other hand, when the wavelengths of perturbations are larger than the horizon, to attach observers to an invisible background is not a convenient choice. In the second case, called the comoving proper-time gauge, the observers are attached instead to the free-falling particles, so they do not see any velocity field (unless there are other non-gravitational forces, like pressure gradients) and, being always free falling, do not measure a gravitational potential. This gauge, therefore, does not have a straightforward Newtonian limit. One can also define quantities that are gauge-invariant, but this does not necessarily simplify the understanding and often leads to cumbersome mathematics. Although a gauge transformation is expressed as a change of coordinates, it is important to observe that (unlike ordinary coordinate transformations) it does not
42
Basics of cosmological perturbation theory
link different observers in the same spacetime but it links two different spacetimes, the background and the perturbed one, seen by the same observer. This is the reason, for instance, why scalar quantities change under a gauge transformation but not under a coordinate transformation. For a detailed discussion of gauge choices, see Refs. [70, 89, 90, 91, 92, 93, 94]. 4.2 The Newtonian gauge Let us start then with the Newtonian gauge. The most general perturbed metric can (0) + δgµν , where be written schematically as gµν = gµν wi 2 −2 . (4.4) δgµν = a wi 2δij + hij Here and are spatial scalars, wi is a 3-vector, and hij is a traceless 3-tensor. All the perturbation quantities (, , wi , etc.) depend on space and time although we will not necessarily indicate this explicitly for simplicity of notation. For instance, it is easy to show that g00 is a spatial scalar. If x˜ µ = f (x µ ) is a general transformation, the metric tensor transforms as g˜ µν =
∂x α ∂x β gαβ . ∂ x˜ µ ∂ x˜ ν
(4.5)
If we perform a purely spatial transformation, x˜ 0 = x 0 , x˜ i = f (x i ), we have immediately that g˜ 00 = g00 , as requested for a spatial scalar. (0) + δgµν , the condition that If we write the perturbed metric as gµν = gµν γβ β gαγ g = δα imposes the following relation at first order: δg µν = −δgαβ g (0)αµ g (0)βν .
(4.6)
That is, the inverse of the perturbed metric is minus the perturbed metric with indices raised by the unperturbed metric. A decomposition analogous to gµν can be done for any rank-two tensor as e.g., the energy-momentum tensor. From Helmholtz’s theorem one can decompose the vector wi into a longitudinal and a transverse component
wi = wi + wi⊥ ,
(4.7)
where by construction
∇ · wi⊥ = ∇ × wi = 0 .
(4.8)
The longitudinal component, wi , being curl-free, is the gradient of a scalar quantity ws , i.e. wi = ∇ws . When we derive the Einstein equations for the (0i) components, we will have therefore longitudinal and transverse terms, both in G0i and in T0i .
4.2 The Newtonian gauge
43
Taking the curl of the equations, we are left with only the transverse equations. On the other hand, taking the divergence, we are left with the longitudinal ones. Therefore, the two components completely decouple from each other and evolve independently, and therefore can be treated separately. Since the density perturbation δ is a scalar quantity, only the longitudinal terms, which can be derived from a scalar quantity, couple to the density perturbations. A similar argument holds for the traceless spatial part hij . This tensor can be written in general as a sum of three traceless terms:
T hij = hij + h⊥ ij + hij ,
(4.9)
where the divergences ∂ i hij , ∂ i h⊥ ij (which are vectors) are longitudinal and transverse, respectively, and hTij is transverse, that is
ij k ∂i ∂k hij = 0,
∂i ∂j h⊥ ij = 0 ,
∂i hTij = 0 .
(4.10)
Here the Levi-Civita tensor ij k is +1 for even index permutations (123,312,231), −1 for odd permutations, and 0 for repeated indices. This means that the curl of T ∂i hij as well as divergences of ∂j h⊥ ij and hij vanish. Now, since ∂i hij is curl-free, it can be written in terms of a scalar function B, and it is easy to verify that ij k ∂i ∂k hij = 0 if 1 2 (4.11) hij = ∂i ∂j − δij ∇ B ≡ Dij B , 3 where the traceless operator Dij is defined implicitly. On the contrary, the perturT bations h⊥ ij , hij cannot be derived from a scalar function. The first one is a vector giving rise to rotational velocity perturbations, whereas the second one is a tensor giving rise to gravitational waves. Both decouple completely from the scalar terms and can be treated separately. The terms which are intrinsically vectorial couple to pure rotational modes, while tensorial terms represent gravitational waves, coupled to matter only for anisotropic perturbations. Furthermore, it can be shown that if initially the rotational, or vorticity, modes are zero, they remain zero throughout (unless there are entropy gradients). If they are present initially, they decrease as a −1 . Since they are not of much interest in current dark energy research we will not further deal with vector or tensor perturbations. Therefore, we need to take into account only the part of wi and hij derived from scalars. This may be done by introducing two new scalar functions, E and B, that produce the vector E,i and the tensor Dij B, in analogy to the electromagnetic forces. Then the perturbed metric is given by E,i 2 −2 . (4.12) δgµν = a E,i 2δij + Dij B
44
Basics of cosmological perturbation theory
Out of the four scalar functions , , E, B, one can construct gauge-invariant quantities, that is, combinations that remain invariant at first-order under a general coordinate infinitesimal transformation, x˜ µ = x µ + ξ µ . As we have already mentioned, the situation can be much simplified if one works in a specific gauge. This can be done by imposing up to four conditions on the metric, which corresponds to the four gauge coordinate transformations. Here we choose them to be wi = 0 (from which E = 0) and B = 0. This finally leaves the perturbed metric in the Newtonian or longitudinal or shear-free gauge: ! ds 2 = a 2 (η) −(1 + 2)dη2 + (1 + 2)δij dx i dx j .
(4.13)
Beware of the signs: many authors choose opposite metric signature and/or opposite signs for and , which can lead to a great deal of confusion. We follow the choice of Dodelson’s textbook [74]. In order to derive the first-order Einstein equations, we decompose the Einstein tensor Gµν and the energy-momentum tensor Tνµ into background and perturbed parts: Gµν = Gµ(0) + δGµν and Tνµ = Tνµ(0) + δTνµ . The background cosmological ν = evolution is obtained by solving the zero-th order Einstein equations, Gµ(0) ν 8π G Tνµ(0) . The first-order Einstein equations are given by δGµν = 8π GδTνµ .
(4.14)
The l.h.s. of Eq. (4.14) can be computed by the following procedure. We first need µ to calculate the perturbed Christoffel symbols δνλ by using the formula: 1 1 µ δνλ = δg µα (gαν,λ + gαλ,ν − gνλ,α ) + g µα (δgαν,λ + δgαλ,ν − δgνλ,α ) . 2 2 (4.15) For the metric (4.13), the non-vanishing components of perturbed Christoffel symbols are ! δij0 = δij 2H ( − ) + ,
(4.16)
0 = , δ00
(4.17)
0 i δ0i = δ00 = ,i ,
(4.18)
δji 0 = δji .
(4.19)
Note that in this chapter a prime represents the derivative with respect to the conformal time η apart from the use for dummy integration variables such as k and r .
4.2 The Newtonian gauge
45
The next step is to derive the perturbations in the Ricci tensor and in the Ricci scalar: β
β
α α α α α β α β δRµν = δµν,α − δµα,ν + δµν αβ + µν δαβ − δµβ αν − µβ δαν ,
(4.20) δR = δg µα Rαµ + g µα δRαµ .
(4.21)
Finally the perturbed Einstein tensors are derived by 1 1 δGµν = δRµν − δgµν R − gµν δR , 2 2 δGµν = δg µα Gαν + g µα δGαν .
(4.22) (4.23)
For the metric (4.13) we obtain
! δG00 = 2a −2 3H(H − ) + ∇ 2 , δG0i = 2a −2 − H |i ,
! δGij = 2a −2 (H2 + 2H ) + H − − 2H δji + a −2 ∇ 2 ( + )δji − ( + )| i j ,
(4.24) (4.25)
(4.26)
where the subscript “| ” represents a covariant derivative with the spatial 3-metric ;µ and ∇ 2 f ≡ f;µ . If the matter source is specified, the perturbed energy-momentum tensor δTνµ is determined accordingly. We then obtain the linear perturbation equations from Eq. (4.14). We also recall that the energy-momentum tensor satisfies the µ continuity equation Tν;µ = 0. The first-order part of this equation, µ = 0, δTν;µ
(4.27)
also gives a number of useful equations, as we will see later. In order to evaluate the perturbed energy-momentum tensor, we need to perturb µ the four-velocity uµ ≡ dxds . Neglecting the perturbations higher than the first order, we obtain vi 1 µ (1 − ) , , u = a a uµ = gµν uν = [−a (1 + ) , avi ] , uµ uµ = −1 , i
i
(4.28)
where v i = dx = a dx is the matter peculiar velocity with respect to the general dη dt expansion. A note about the definition of the peculiar velocity is in order here. From Eq. (2.58) the comoving distance at redshift z is given by (after setting c = 1
46
Basics of cosmological perturbation theory
and a0 = 1)
dc (z) = 0
z
d˜z . H (˜z)
(4.29)
However, due to the peculiar velocity of the source, the redshift is the sum of cosmological expansion and peculiar velocity Doppler effect, vp z = zc + zp = zc + (zp zc ) , (4.30) a where vp = azp is the projection of the peculiar velocity vector along the line of sight. Then we have zc zc zc +zp zp (1 + zc )vp d˜z d˜z d˜z dcˆ (z) = + = + . (4.31) H (˜z) z) H (zc ) z) H (zc ) 0 0 H (˜ 0 H (˜ The extra apparent distance induced by the peculiar velocity is therefore (1 + z)H (z)−1 vp . This can be always neglected, except when one is interested in the properties of the redshift space/real space conversion, see Section 14.3.
4.3 Single-fluid model Let us consider a single-fluid model with an energy-momentum tensor Tµν . For a general fluid the energy momentum tensor is given by ! Tµν = (ρ + P )uµ uν + P gµν + qµ uν + qν uµ + πµν , (4.32) where, beside the familiar symbols ρ, P , uµ for the energy density, the pressure and the four-velocity vector, we meet the heat flux vector qµ and the viscous shear tensor πµν . The terms inside square brackets in Eq. (4.32) are important only for fluids whose internal energy is a sizable fraction of the total energy density. We have qµ = 0 and πµν = 0 for perfect fluids. In the following we limit ourselves to perfect fluids. We also assume that the perturbed fluid remains a perfect fluid. This implies that ji ≡ δTji = 0 (i = j ), a condition that will be used below. The notation for the perturbed quantities is δ≡
δρ , ρ
θ ≡ ∇i v i ,
(4.33)
where δρ/ρ ≡ (ρ(x) − ρ)/ ¯ ρ¯ is the density contrast (ρ(x) is the density field and ρ¯ is the spatial average) and θ is the velocity divergence. In general there are several pairs δi , θi , one for each perfect fluid composing the Universe. All of the perturbed quantities are functions of space x and time t. A simple observation can avoid much confusion later on: we will often speak loosely of δ(x) as if it were a deterministic quantity and we will say that δ grows
4.3 Single-fluid model
47
or decays. However δ(x) is in reality a random field which by definition has a zero mean value δ = 0. What we mean is that in the linear regime the value of δ(x) at any point x grows or decays so that δ(x, t) = D(t)δ(x, 0), where D(t) is the growth (or decay) function. In the linear regime the spatial part is always factored out and its properties are assigned by initial conditions. We will always assume Gaussian initial conditions as predicted in standard inflationary models. From Eq. (4.32) the perturbed energy-momentum tensor for a perfect fluid with the equation of state w = P /ρ can be written as δTνµ = ρ[δ(1 + cs2 )uν uµ + (1 + w)(δuν uµ + uν δuµ ) + cs2 δ δνµ ] ,
(4.34)
where δνµ should not be confused with the density contrast δ. Here we have introduced the sound velocity, cs2 ≡ δP /δρ. If P , even when perturbed, depends on ρ alone (which is the case called barotropic fluid) then cs2 ≡
P˙ δP dP = = . δρ dρ ρ˙
(4.35)
The last passage is valid only in the FLRW metric where at background level everything depends on time alone (cs is calculated at zero-th order since it will always appear as a factor of first-order variables). Since cs , just as w, depends at first-order only on background quantities, in this case the perturbation equations do not introduce any new free function. In general, however, the pressure P can depend on internal degrees of freedom of the fluid, say, entropy s. Then one has cs2 =
δP (ρ, s) ∂P ∂P ∂s 2 2 + cs(na) , = + = cs(a) δρ ∂ρ ∂s ∂ρ
(4.36)
where cs(a) ≡ P˙ /ρ˙ is called the adiabatic sound speed and cs(na) is, guess what, the non-adiabatic sound speed. The non-adiabatic sound speed in general will depend on microphysical properties of the fluid and appears as a new free function only at the level of perturbations. The gravitational equations at first-order are then completely specified only if we give for each fluid the equation of state w(a) and the total sound speed cs (a) or, equivalently, if we assign to the fluid a function P (ρ, s) which determines both. The components of the energy-momentum tensor are δT00 = −δρ ,
(4.37)
δTi0 = −δT0i = (1 + w)ρv i ,
(4.38)
δT11 = δT22 = δT33 = cs2 δρ .
(4.39)
48
Basics of cosmological perturbation theory
Then the perturbed Einstein equations (4.14) lead to 3H(H − ) + ∇ 2 = −4π Ga 2 δρ ,
(4.40)
∇ 2 ( − H) = 4π Ga 2 (1 + w)ρθ ,
(4.41)
= − ,
(4.42)
+ 2H − H − (H2 + 2H ) = −4π Ga 2 cs2 δρ .
(4.43)
Note that Eqs. (4.40)–(4.43) come from the (00), (0i), (ij ), and (ii) components. Equation (4.42) follows from the property δTji = 0. One can also derive some useful equations by using the continuity equation (4.27). Recall that the operation of covariant divergence of a tensor Tνµ is µ µ α α = Tν,µ − νβ Tαβ + βα Tνβ . Tν;µ
(4.44)
µ
Then the ν = 0 component of Eq. (4.27), i.e. δT0;µ = 0, reads β
µ
α α α α Tαβ − 0β δTαβ + δ0α T00 + βα δT0 = 0 , δT0,µ − δ0β
(4.45)
which reduces to (δρ) + 3H(δρ + δP ) = −(ρ + P )(θ + 3 ) ,
(4.46)
where we have employed Eqs. (4.16)–(4.19). Using the unperturbed conservation equation ρ + 3H(ρ + P ) = 0 together with the relations w = P /ρ and cs2 = δP /δρ, we find that Eq. (4.46) can be expressed as δ + 3H(cs2 − w)δ = −(1 + w)(θ + 3 ) ,
(4.47)
which is called the (perturbed) continuity equation. For non-relativistic matter with w = 0 and cs2 = 0, this equation reduces to δ = −θ − 3
(for non-relativistic matter).
(4.48)
This equation tells us that the density at position x increases if there is a velocity divergence in the same place, that is, if there is more matter coming in than going out. The term, absent in Newtonian dynamics, is negligible at small scales and, of course, for a slowly varying gravitational potential. µ The equation δTν;µ = 0 for ν = i leads to δq + 3Hδq = −aδP − (ρ + P )a ,
(4.49)
where δq ≡ a(ρ + P )v and v is a velocity potential related to v i via v i = ∇ i v. Writing Eq. (4.49) in terms of v i and taking the divergence ∇i , we obtain 2 cs w 2 θ = −∇ δ+ . (4.50) θ + H(1 − 3w) + 1+w 1+w
4.3 Single-fluid model
49
For non-relativistic matter, this reduces to θ + Hθ = −∇ 2 − ∇ 2 (cs2 δ)
(for non-relativistic matter) ,
(4.51)
where now we have included the ∇ 2 (cs2 δ) term. This is called the Euler equation in the Newtonian context. It says that the (peculiar) acceleration depends on the sum of the potential and pressure gradients. We go now to the Fourier space. This means that all perturbation quantities will be Fourier expanded (we do not need to bother with pre-factors here): ik·r 3 = e k d k, = eik·r k d3 k , (4.52) ik·r 3 (4.53) δ = e δk d k , θ = eik·r θk d3 k . The subscript k represents a Fourier mode for each wavenumber k. Note that k is a comoving quantity that remains fixed. In the following we drop the subscript k as long as no confusion arises by doing so. In Fourier space we assume that the perturbation variables (δ, θ, , etc.) are the sum of plane waves δk eik·r . Since the equations are linear, each plane wave obeys the same equations with a different comoving wavenumber k. Throughout the linear evolution, the physical scale λp of the perturbation expands with the cosmic expansion as λp = (2π/k)a. Of course, if the perturbation enters a non-linear regime, then this treatment breaks down and the perturbation decouples from the Hubble expansion and starts collapsing. When we calculate the perturbation equations with an algebraic manipulator it is very convenient to introduce from the beginning all perturbation variables as Fourier modes, e.g., δ(x, y, z, t) = δk (t)eik·r . Since we are always interested in the direction-averaged equations (i.e. the equations √that depend only on the modulus k), we could simply put k · r = k(x + y + z)/ 3. In practice, each perturbation quantity φ and its derivatives can be substituted as follows φ(x, η) → eik·r φ(η) ,
(4.54)
∇φ(x, η) → ieik·r k φ(η) ,
(4.55)
∇ 2 φ(x, η) ≡ ∇i ∇ i φ(x, η) → −eik·r k 2 φ(η) .
(4.56)
Notice that when there are two repeated spatial indices we sum over them without the help of the metric coefficients gij (more exactly, we use the induced 3-dimensional spatial metric which for spatially flat spaces is just the Euclidean metric). Furthermore, the Fourier modes eik·r can be simply dropped out, since the equations are linear and therefore decoupled between different modes.
50
Basics of cosmological perturbation theory
From Eqs. (4.40)–(4.43), (4.47), and (4.50) we obtain the following equations for each Fourier mode: k 2 + 3H( − H) = 4π Ga 2 ρδ ,
(4.57)
k 2 ( − H) = −4π Ga 2 (1 + w)ρθ ,
(4.58)
= − ,
(4.59)
+ 2H − H − (H2 + 2H ) = −4π Ga 2 cs2 ρδ ,
(4.60)
δ + 3H(cs2 − w)δ = −(1 + w)(θ + 3 ) , 2 cs w 2 θ =k δ+ , θ + H(1 − 3w) + 1+w 1+w
(4.61)
θ = ik · v .
(4.63)
(4.62)
where now
Although the six equations above are not independent they are all useful. Let us remark again that we are considering here the fluctuations of a Universe composed of a single fluid. In the next sections we will solve the general set of equations (4.57)–(4.62) in several different regimes: during the epochs of radiation and matter eras, and both at large and small scales. Finally, one can combine Eqs. (4.57) and (4.58) to get the relativistic Poisson equation k 2 = 4π Ga 2 ρ[δ + 3H(w + 1)θ/k 2 ] = 4π Ga 2 ρδ ∗ ,
(4.64)
where we define the total-matter variable: δ ∗ ≡ δ + 3H(w + 1)θ/k 2 .
(4.65)
Notice that in our conventions an overdensity δ ∗ > 0 generates opposite gravitational potentials, > 0 and < 0, on small scales. Combining Eqs. (4.57), (4.59), and (4.60), we can get an equation for alone: + 3H(1 + cs2 ) + (cs2 k 2 + 3H2 cs2 + 2H + H2 ) = 0 .
(4.66)
Similarly, using the relativistic Poisson equation (4.64), this becomes an equation for δ ∗ : 3 2 ∗ 2 ∗ 2 2 2 2 (δ ) + H(1 + 3cs − 6w)(δ ) − H (1 − 6cs − 3w + 8w) − cs k δ ∗ = 0 , 2 (4.67)
4.5 Scales smaller than the Hubble radius
51
where cs , w are arbitrary functions of time and where we have employed the useful relation 1 H = − (1 + 3w)H2 . (4.68) 2 4.4 Scales larger than the horizon Now that we have derived the perturbation equations, we can begin wondering how to solve them. As a first example we work out the simplest case, the large-scale limit k H = aH . This corresponds to the scale on which the physical wavelength λp = (2π/k)a of perturbations is much larger than the Hubble radius H −1 , i.e. super-horizon scales (although notice that the horizon corresponds approximately to 1/H only for some particular case). If the pressure depends only on the energy density and the equation of state w is a constant then we have cs2 = w, which is valid both for matter and radiation. In this case Eq. (4.66), using Eq. (4.68), reduces to + 3H(1 + cs2 ) = 0 .
(4.69)
Then = 0 is a solution. Equation (4.57) becomes 3H2 = 4π Ga 2 ρδ .
(4.70)
Using the Friedmann equation, 3H2 = 8π Gρa 2 , it follows that δ = 2 .
(4.71)
Hence = constant at large scales implies that δ = constant. One easily finds that the result (4.71) is consistent with the other Einstein equations. Equation (4.69) is second-order, so we must have two solutions. It appears immediately that = constant is a growing mode or a dominating solution (at least for cs2 > −1). Thus we have shown that the gravitational potential remains constant for scales outside the Hubble radius whenever cs2 = w for the total fluid. During the transition from radiation to matter eras this condition is violated and as we will see in Section 4.10 the gravitational potential changes. 4.5 Scales smaller than the Hubble radius Next, we work out the opposite case, k H, i.e. scales deep inside the Hubble radius (sub-horizon scales). The fluctuation of a pressureless fluid can grow indefinitely because there is no counteracting force. In general, however, the pressure of the fluid resists gravity and stops the collapse. It is then instructive to derive the equations for a fluid which is pressureless (w = 0) in the absence of perturbations,
52
Basics of cosmological perturbation theory
but has a small sound speed: cs2 =
δP 1. δρ
(4.72)
Then Eq. (4.58) tells us that − H 0, so that Eq. (4.57) corresponds to the Fourier transformed Poisson equation 3 (4.73) k 2 = 4π Ga 2 ρδ = H2 δ . 2 Taking the derivative of Eq. (4.73) and substituting it into Eq. (4.61), we obtain H 9 H2 δ δ = −θ − −θ . (4.74) δ 2 + 2 k2 H δ Hence this equation reduces to the energy conservation equation in the Newtonian limit. Then the perturbation equations in the sub-horizon limit become δ = −θ ,
(4.75)
θ = −Hθ + cs2 k 2 δ − k 2 ,
(4.76)
and Eq. (4.73). Differentiating Eq. (4.75) with respect to η and using Eq. (4.76), it follows that 3 2 2 2 (4.77) δ + H δ + cs k − H δ = 0 . 2 In the Minkowski limit, H → 0, this equation reduces to the classical fluid wave equation δ + cs2 k 2 δ = 0, where cs is indeed the sound velocity. Equation (4.77) shows at once that the perturbation does not grow if 3 (4.78) cs2 k 2 − H2 > 0 , 2 i.e. if the physical wavelength λp = (2π/k)a is smaller than the Jeans length, " π . (4.79) λJ = cs Gρ For scales smaller than λJ the perturbations undergo damped oscillations. For the CDM particles the velocity dispersion is always negligible, at least√in the regime of validity of our linear treatment. For the photons we have cs = c/ 3, so that λJ ≈ H −1 .
(4.80)
Hence the growth of perturbations is prevented on all scales smaller than the Hubble radius. For the baryons, the sound velocity is comparable to the photon
4.6 Two-fluid solutions
53
velocity before the decoupling epoch, so that baryon perturbations are damped out (more precisely they drop rapidly to a comoving scale of less than 1 Mpc just after decoupling). Then the baryons are free to fall inside the dark matter potential wells, and their perturbation spectrum catches the dark matter one (more on this in the next section). When cs k H, the perturbations grow freely because gravity overcomes the pressure: this is the all-important regime of gravitational instability. The subhorizon equation for a single pressureless fluid becomes 3 δ + Hδ − H2 δ = 0 . (4.81) 2 It is often useful to employ the number of e-foldings N = ln a. Then Eq. (4.81) can be written as d2 δ 1 dH 3 dδ + +1 − δ = 0. (4.82) 2 dN H dN dN 2 We can rewrite Eq. (4.68) as 1 dH 1 3 = − − w. H dN 2 2 For a pressureless fluid (w = 0), Eq. (4.82) then reduces to d2 δ 3 1 dδ − δ = 0. + 2 dN 2 dN 2
(4.83)
(4.84)
The direct substitution δ = AeλN gives the solution λ = 1, −3/2. Then the evolution of growing and decaying modes during the matter era is given by δ+ = Aa,
δ− = Ba −3/2 .
(4.85)
In terms of the cosmic time t, the growing solution evolves as δ+ ∝ t 2/3 . The pre-factor is of course fixed by the initial conditions, ultimately established during inflation. The decaying solutions (or in general the slower one) become soon negligible with respect to the growing ones and we will systematically neglect them throughout this book. Inserting δ+ into the Poisson equation (4.73), we see that ∝ a 2 H 2 δ+ ∝ a 2 a −3 a 1 ∝ constant. Hence the gravitational potential remains constant during the pure matter-dominated epoch.
4.6 Two-fluid solutions We generalize the single-fluid case to the more realistic case in which both matter (wm = cs2 = 0) and radiation (wr = cs2 = 1/3) are present. We introduce the matter
54
Basics of cosmological perturbation theory
perturbation variables δm , θm and the radiation perturbation variables δr , θr (here radiation means all the components which are massless or relativistic, in particular the neutrinos). Since we are considering dark matter as a dominant matter component, there is no explicit interaction term between matter and radiation. The baryonic fraction is also effectively decoupled after z ≈ 1000, while before this epoch it can be considered as a part of a relativistic photon–baryon plasma. In Fourier space we then have a system of gravitationally coupled equations for the perturbations on sub-horizon scales: = −(θm + 3 ) , δm
(4.86)
θm = −Hθm − k 2 ,
(4.87)
4 δr = − (θr + 3 ) , 3 2 3 2 c δr − , θr = k 4 s
(4.88) (4.89)
k 2 ( + H) = −4π G(1 + weff )a 2 ρt θt ,
k + 3H( + H) = 4π Ga ρt δt . 2
2
(4.90) (4.91)
The subscript t represents total perturbation variables, i.e. ρt = ρm + ρr ,
(4.92)
ρr /3 , ρm + ρr (1 + wm )m θm + (1 + wr )r θr , θt = 1 + weff δt = m δm + r δr .
weff = r wr + m wm =
(4.93) (4.94) (4.95)
Here the total effective equation of state weff = Pt /ρt is given by weff = −1 −
2 H˙ , 3 H2
(4.96)
which follows from Eqs. (2.17)–(2.18) and K = 0. We remind the reader that m and r are functions of time. These should be (0) distinguished from their present values (0) m and r . In the sub-horizon limit, Eq. (4.91) gives 3 k 2 4π Ga 2 (ρm δm + ρr δr ) = H2 (m δm + r ρr ) . 2
(4.97)
4.6 Two-fluid solutions
55
Following the derivation similar to Eq. (4.77), we obtain the following equations for sub-horizon perturbations 3 + H δm − H2 (m δm + r δr ) = 0 , δm 2 2 k δr + δr = 0 . 3
(4.98) (4.99)
During the radiation-dominated epoch we have m 0 and r 1. Moreover the second equation shows that the radiation density contrast oscillates rapidly around zero (since we are considering sub-horizon modes, k H). The same is true for the coupled baryon–photon plasma. Therefore, we can average over the radiation oscillations and put δr 0 in the first equation. It then follows that m δm + r δr 0 and δm + H δm 0.
(4.100)
The solution of this equation is given by δm = C1 + C2 dη/a. During the radi ation era the integral dη/a gives only a logarithmic correction, so the matter perturbations evolve only mildly. During the matter era we have |m δm | |r δr | in Eq. (4.98), so that the evolution of matter perturbations is described by δm ∝ a as we have explained in the previous section. If we consider cold dark matter (perturbation δc ) and baryonic matter (perturbation δb ) instead of matter and radiation, Eq. (4.98) can be generalized as 3 δc + Hδc − H2 (c δc + b δb ) = 0 , 2 3 δb + Hδb − H2 (c δc + b δb ) = 0 . 2
(4.101) (4.102)
Since baryons correspond to a small fraction of the total matter fluid, we can assume |b δb | |c δc |. This shows that Eq. (4.101) decouples from δb and reduces to the standard equation for matter perturbations. At the same time the baryon equation is “forced” by the term c δc . For such coupled differential equations the asymptotic solution of δb will approach the forcing term δc . In other words, the perturbations in baryons will catch up with those in dark matter. This expresses mathematically (in the linear regime) the common expression according to which the baryons fall into the dark matter potential wells. Analogously, if we consider the sum of pressureless matter and the cosmological constant instead of matter and radiation, we get the term δ in addition to m δm . However ρ is constant by definition and δ = 0, so that we have a slight
56
Basics of cosmological perturbation theory
modification of Eq. (4.81): 3 + H δm − H2 m δm = 0 . δm 2 This equation can be rewritten in terms of the derivative N : dδm 3 d2 δm 1 dH +1 − m δm = 0 . + 2 dN H dN dN 2
(4.103)
(4.104)
If we assume that m = constant, then the solution is given by δm ∼ a λ± with
1 (4.105) −1 ± 1 + 24m . λ± = 4 This case occurs when the fraction 1 − m is into some form of energy density which has w ≈ 0 but contrary to ordinary CDM it does not cluster on sub-horizon scales. The major example of this is massive neutrinos after they become nonrelativistic. Supposing for a moment we could apply it also for CDM, this would show that the cosmological constant slows down the perturbation growth. In the limit m → 0 we have λ+ → 0 from Eq. (4.105), which is qualitatively correct. However, the density parameter m =
ρm ρ (0) a −3 = (0) m ρm + ρ ρm a −3 + ρ
(4.106)
is obviously not a constant. A much better approximation, obtained by an empirical fit, is given by defining the growth rate f of matter perturbations [95]: f ≡
d ln δm = γm , d ln a
that is
δm (a) = δm (ai ) exp
a
(4.107)
da˜ ˜ m (a) a˜ γ
ai
,
(4.108)
where the growth index γ is ≈ 0.55 for the CDM model (see Section 11.1 for the derivation and for a generalization to dark energy). With this behavior we realize that the term H2 δm in the Poisson equation is no longer constant and therefore the gravitational potential on sub-horizon scales is not constant. For the CDM model the gravitational potential is almost constant during the matter era, but it begins to decrease after the universe enters the dark-energy-dominated epoch. 4.7 Velocity field The mass power spectrum can be studied also by analyzing the peculiar motion of the galaxies. It is intuitive, in fact, that a more clustered distribution of matter will
4.7 Velocity field
57
induce stronger peculiar velocities. The importance of this is that the velocity field depends on the total mass distribution, including any unseen component. Let us start from Eq. (4.75) with θ = ∇i v i . In Fourier space this can be written as δk = −iki v i .
(4.109)
In this section we recover the subscript k for Fourier modes. This equation applies to each pressureless component such as baryons and CDM separately. However, due to the gravitational coupling, each component will be driven by a dominating density contrast, as we have seen explicitly for the baryons. We also assume that the velocity field v can be represented by the galaxy velocity field, v g . This of course assumes that the galaxy velocities are not biased with respect to the dark matter velocity field, an expectation based on the fact that galaxies move because of the common gravitational field. Strictly speaking, this only implies that the acceleration, not the velocity, is the same, and only if all matter components share the same equation of state and sound speed. Assuming similar initial conditions, universality of the gravitational interaction, and identical equation of state and sound speed, however, the galaxies can be seen as test particles of a universal peculiar velocity field. Under these assumptions, observing the peculiar velocity field v g of galaxies gives information on the total density contrast. So in this section we take v to refer to the velocity field of galaxy “test particles” and δk to refer to the total mass. Consider Eq. (4.51) for cs = 0:
v i = −Hv i + ik i k .
(4.110)
Since we are dealing only with scalar perturbations, the velocity can be written as the gradient of a velocity potential v, i.e. v i = ∇ i v → ik i v. Then it is clear that v i is parallel to k i and we can look for solutions of Eq. (4.109) in the form v i = F (k, a)k i . This gives immediately from Eq. (4.109) the relation between the peculiar velocity field v i and the density fluctuation δk in linear perturbation theory (in the Newtonian regime): v i = i H f δk
ki , k2
(4.111)
where f is the growth rate defined in Eq. (4.107). Substituting Eq. (4.111) into Eq. (4.109), one can easily confirm that the relation f = δk /(Hδk ) = d ln δk /d ln a follows. During the standard matter-dominated era we have already seen that δk ∝ a and hence f = 1 while more in general f = m (a)γ .
58
Basics of cosmological perturbation theory
Let us consider the present epoch (and motion on a local scale, say, less than 100 Mpc). Since a = a0 = 1 and hence H = H0 , we have k . (4.112) k2 The peculiar velocity v(r) at location r can be obtained by Fourier antitransformation: k V v(x) = iH0 f (4.113) δk 2 eik·r d3 k . 3 k (2π ) v = iH0 f δk
Here we implicitly assumed f to be k-independent. As we have seen this is true in CDM but not necessarily in other models. The average in a volume VR of radius R is k ik·r k V V 3 3 δ δk 2 W (kR)d3 k . v R = iH0 f e W (r)d k d r = −iH f k 2 0 3 3 k k (2π ) VR (2π ) (4.114) Here W (kR) is the Fourier transform of the window function, defined as 1 (4.115) W (r)e−ik·r d3 r . W (kR) ≡ VR Therefore, the average of the square of the velocity is 2 2 2 2 V ∗ k k δk δk 2 2 W (kR)W (k R)d3 kd3 k v R = H0 f 6 k k (2π ) H 2f 2 k k = 0 3 P (k)δD (k − k ) 2 2 W (kR)W (k R)d3 kd3 k k k (2π ) H 2f 2 = 0 2 (4.116) P (k)W 2 (kR)dk , 2π where in the second line we have used the definition in Eq. (3.26) and in the last line we integrated over the solid angle 4π . The square root of v 2 R is the magnitude of the peculiar flow on the scale R and is called bulk flow. Estimates of the bulk flow can be used to constrain or normalize the mass power spectrum (see e.g., Refs. [96, 97]). On the other hand, independent measures of P(k) and v 2 R can 2γ give f 2 m . 4.8 The redshift distortion The galaxy distances are mostly measured through their redshifts. However, the redshift includes the peculiar velocity of the galaxies themselves, so that there is an error in the distances we assign to galaxies. On very small scales, i.e. in the
4.8 The redshift distortion
59
cluster cores, the peculiar velocity of a galaxy is more or less randomly oriented, so that the error in the distance is statistical, and can be taken into account along with the experimental errors. On redshift maps, the small scale peculiar velocities cause the fingers-of-god effect: galaxies in a cluster acquire an additional random redshift that distorts the cluster distribution, making it appear elongated along the line of sight. On large scales, however, the galaxies tend to fall toward concentrations, so that the velocity field is coupled to the density field. This correction is systematic and can be accounted for in the following way [98]. Given a peculiar velocity v of a source at position r, one can define the line-ofsight component u(r) ≡ v ·
r , r
(4.117)
where r = |r|. The coordinate transformation from real space (r) to redshift space (s) is given by u(r) − u(0) . s = r 1+ r
(4.118)
Here we express velocities in Megaparsecs through division by H0 . If dVs and dVr are the volume elements in the two coordinates with number densities n(s) and n(r), respectively, we have n(r)dVr = n(s)dVs .
(4.119)
The volume element dVs can be written in terms of the r coordinate as u(r) 2 u(r) 2 |J | (r 2 sin θ )drdθ dφ = 1 + |J | dVr , (4.120) dVs = 1 + r r where u(r) ≡ u(r) − u(0). The Jacobian |J | is given by the derivative of s = r[1 + (u(r) − u(0))/r] with respect to r: # # # ∂s # du . |J | = ## ## = 1 + ∂r dr
(4.121)
Using the average density n0 , the density contrast in s-space is δs =
n(s)dVs n(r)dVr n(r) −1= −1= − 1, n0 dVs n0 dVs n0 (1 + u(r)/r)2 |J |
(4.122)
60
Basics of cosmological perturbation theory
where we have used Eq. (4.120). To first order, this yields u(r) du n(r) −1 1−2 − δs n0 r dr = δr − 2
u(r) du − , r dr
(4.123)
where in the last line we have employed the fact that to first-order we can approximate n(r) to be n0 . Therefore, we see that the density contrast is different in the two spaces. As a consequence, the correlation function and the power spectrum measured in redshift space will have to be corrected to be expressed in real space. To do so, we have to take the velocity field from the linear perturbation theory. However what we observe is the galaxy density contrast δg , which is different from the total density contrast δm . The two quantities can be assumed to be related by a bias factor b defined by b≡
δg , δm
(4.124)
so that instead of f δk we have δ(g)k f/b in Eq. (4.112). In other words, wherever we write f , we should in fact write β ≡ f/b. If we assume b to be scale-independent, then Eq. (4.113) can be written as k (4.125) v = iH0 β δ(g)k eik·r 2 d3 k ∗ , k where the Fourier factor V / (2π )3 is included in the differential d3 k ∗ . Notice the apparent oddity: in this equation we have both the total δ (in β) and the galaxy δ(g) . In general however b can depend on scale and time. Notice that if b depends on space then Eq. (4.124) in real space is different from the analogous relation in Fourier space. An immediate consequence of Eq. (4.124) written in Fourier space is that for the power spectra we have Pg (k) = b2 Pm (k) .
(4.126)
One should take note however what this standard procedure is assuming. If linear gravity were the only force at work, then any bias in δ should induce a bias in v due to the continuity equation, while here we are implicitly assuming that biased galaxies possess an unbiased velocity field. In other words, a value b = 1 implies that galaxies violate the linear continuity equation, while dark matter still obeys it. This could only occur if some additional force acts on baryons but not on dark matter. If gravity does not violate the equivalence principle (see Section 11.2 for an example of this), one needs to invoke some non-gravitational effect, e.g.,
4.8 The redshift distortion
61
the hydrodynamical effect, that makes the galaxies “appear” or “disappear” and therefore breaks continuity. Examples of these are merging processes or evolutionary processes that render galaxies brighter or dimmer and therefore visible or invisible to our telescopes. We can say that the whole idea of a simple biasing scheme tries to capture glimpses of physics beyond the purely linear gravitational treatment. Ultimately, its validity can only be judged against N -body simulations and observations. In problem 4.1 you can work out an example of constant bias. Using Eq. (4.125), the line-of-sight component (4.117) in Megaparsec units (H0 = 1) is (we drop now the subscript g in δ but we add a subscript r to remind ourselves that this quantity is in real space) k·r u(r) = iβ δrk eik·r 2 d3 k ∗ , (4.127) k r while its derivative is du = −β dr
δrk e
ik·r
k·r kr
2
d3 k ∗ ,
(4.128)
where we have used the relation d ik·r k·r ik·r =i e e . dr r Finally, from Eq. (4.123), we have k·r 2 3 ∗ du ik·r δ s = δr − = δr + β δrk e dk , dr kr
(4.129)
(4.130)
where we have neglected the second term in (4.123) because it is negligible for large r. Now, we can multiply Eq. (4.130) on both sides by V −1 e−ik ·r d3 r and integrate it. We then obtain from Eq. (3.14) the Fourier transform of (4.130) as (4.131) δsk = δrk + β δrk I (k, k )d3 k , (we switched k, k ) where
−3
I (k, k ) = (2π )
e
i(k −k)·r
k · r kr
2 d3 r .
(4.132)
The redshift distortion then introduces a mode–mode coupling. This coupling can be broken in the useful limit of surveys of very small angular scales. In fact, if we can assume that the cosine µ=
k·r kr
(4.133)
62
Basics of cosmological perturbation theory
is almost constant (that is, the survey spans a small solid angle), then we have I (k, k ) = µ2 δD (k − k) and δsk = δrk (1 + βµ2 ) .
(4.134)
The power spectrum defined in Eq. (3.19) reads 2 Ps (k) = V δrk (1 + βµ2 )2 = Pr (k)(1 + βµ2 )2 .
(4.135)
If we average it over angles, we get (4.136) Ps (k) = Pr (k)(1 + 2βµ2 + β 2 µ4 ) , π 2π 1 where the average f (µ) = (4π )−1 0 f (µ) sin θdθ 0 dφ = (1/2) −1 f (µ)dµ gives 1 1 2 1 1 1 4 1 2 4 µ = (4.137) µ dµ = , µ dµ = . µ = 2 −1 3 2 −1 5 We then obtain [98] Ps (k) = Pr (k)(1 + 2β/3 + β 2 /5) .
(4.138)
The power spectrum is then boosted in redshift space, because the velocity field is directed toward mass concentrations. As a result, galaxies seem more concentrated when seen in redshift space. Remember that we have dropped the subscript g but both Pr and Ps refer to the galaxies. As we anticipated, at very small scales, on the other hand, the velocity orientation can be assumed to be random. Then, on these scales, the net effect is that galaxies in redshift space seem distributed over a larger volume and the power spectrum is therefore decreased. Empirical studies have shown that an approximation valid both at small and large scales is given by the following correction [99]: Ps (k) = Pr (k)G(β, y) ,
(4.139)
where e−y 2 π 1/2 erf(y) 2 2 4 (3β + 4βy + 4y ) − [β (3 + 2y 2 ) + 4βy 2 ] , G(β, y) = 5 4 8 y 4y (4.140) −1 and y = kσv H0 with σv being the cloud velocity dispersion along the line of sight. On small scales the effect is to change the slope by a factor k −1 , whereas on large scales the effect is to raise the amplitude as in Eq. (4.138). 2
4.9 Baryons, photons, and neutrinos So far we have considered only perfect fluids which do not possess anisotropic shear (i.e. Tji = 0) and do not interact except gravitationally. Because of this,
4.9 Baryons, photons, and neutrinos
63
their perturbation dynamics has been completely described by functions of the wavenumber modulus k alone, the density contrast, and the gradient of velocity field. However, the cosmic mix contains imperfect fluids for which such a description is insufficient. When the interaction terms or the energy-momentum tensor depend on the full momentum vector P, the fluid needs to be described by its full distribution function f ( P, x, t). Radiation, baryons, and neutrinos are such fluids and in the following we discuss their full perturbation equations. This section is only meant to summarize the main results that will be used elsewhere. We urge the reader to familiarize themselves with the full treatment in standard cosmology textbooks (e.g., [74], whose notation we follow closely). We shall take the perturbed metric (4.13) with the cosmic time t = adη, i.e. ds 2 = −(1 + 2)dt 2 + a 2 (t)(1 + 2)δij dx i dx j .
(4.141)
Given a distribution function f ( P, x, t), the energy-momentum tensor for a fluid in the full general relativistic framework is √ P µ Pν gi dP dP f( P, x, t) , (4.142) dP Tνµ (x, t) = 1 2 3 −g (2π )3 P0 where P 0 ≡ dt/dλs , P i ≡ dx i /dλs (λs is a parameter that characterizes the particle’s path), and gi are the internal degrees of freedom. For a particle of mass m one has P µ Pµ = −m2 . From the expression (4.142) we can derive the usual energy density −T00 and pressure Tii /3. We see that the term that contributes to the shear Tji is proportional to dx i dxj P i Pj = = v i vj , (P 0 )2 dt 2
(4.143)
which is a second-order quantity and therefore is negligible in linear perturbation theory. For massless particles, however, this is not the case. Photons and neutrinos (massless or massive but relativistic) therefore contribute to the shear term. Because of the mass constraint on P µ there are only three momentum degrees of freedom. We can choose the spatial magnitude p 2 ≡ gij P i P j ,
(4.144)
pˆ i ≡ P i /|P | ,
(4.145)
and the unit direction vector
such that δij pˆ i pˆ j = 1. Plugging P i = |P |pˆ i into Eq. (4.144), we find p 2 = a 2 (1 + 2)(δij pˆ i pˆ j )P 2 = a 2 (1 + 2)P 2 ,
(4.146)
64
Basics of cosmological perturbation theory
which gives |P | = p(1 − )/a at first-order. Hence the spatial vector P i can be written as Pi =
1− i p pˆ . a
(4.147)
The photons satisfy the relation gµν P µ P ν = 0, which translates into the following condition − (1 + 2)(P 0 )2 + p2 = 0 .
(4.148)
We then obtain the time-component of P µ : P 0 = p(1 − ) .
(4.149)
The process of collisions between particles can be described by the Boltzmann equation df ∂f ∂f dx i ∂f dp ∂f dpˆ i = + i + + dt ∂t ∂x dt ∂p dt ∂ pˆ i dt = C[f ] ,
(4.150) (4.151)
where f (p, pˆ i , x i , t) is the distribution function and C[f ] describes a collision term. The r.h.s. of Eq. (4.150) should be evaluated up to the first-order for the line element (4.141). The last term in Eq. (4.150) vanishes at the linear level, since both ∂f/∂ pˆ i and dpˆ i /dt are first-order terms. The collision term in Eq. (4.151) is different depending on the kind of matter species. For photons we need to compute it for the process of Compton interaction between protons and electrons (i.e. baryons). Dark matter interacts very weakly with other particles so that the collision term is set to zero. Let us consider the Boltzmann equation (4.151) for the photons. From Eqs. (4.147) and (4.149) it follows that dx i Pi 1−+ i = 0 = pˆ . dt P a
(4.152)
0 From the time component of the geodesic equation dP 0 /dλs = −µν P µ P ν , we obtain the following relation [74] ∂ pˆ i ∂ dp . (4.153) = −p H + + dt ∂t a ∂x i
Then the l.h.s. of the Boltzmann equation (4.151) yields ∂ pˆ i ∂ df ∂f ∂f pˆ i ∂f H+ . −p = + + dt ∂t a ∂x i ∂p ∂t a ∂x i
(4.154)
4.9 Baryons, photons, and neutrinos
65
Note that we have neglected the product of ∂f/∂x i and () because ∂f/∂x i is a first-order term (f does not depend on x i at zero-th order). Recall that in the unperturbed background the photons with the temperature T obey the Bose–Einstein distribution function !−1 f (0) (t, p) = exp(p/T ) − 1 , (4.155) where we have neglected the chemical potential µ. Note that in this background the temperature T depends on the time t only: T ∝ 1/a(t). In the perturbed Universe we can define the temperature perturbation: (t, x, pˆ i ) ≡ δT /T ,
(4.156)
which is chosen as a perturbation variable of photons instead of δγ = 4δT /T (recall that ρr ∝ ργ ∝ T 4 ). We assume that does not depend on the magnitude p since in a Compton scattering this is approximately conserved. The distribution function is given by −1 p ˆ = exp −1 . f (t, p, x, p) T (t)[1 + (t, x, pˆ i )]
(4.157)
If 1, one can expand this about the background value f (0) by using the relation T ∂f (0) /∂T = −p∂f (0) /∂p: f = f (0) − p
∂f (0) , ∂p
(4.158)
which is valid at first-order. Plugging Eq. (4.158) into Eq. (4.154) and collecting the first-order terms, we obtain ∂f (0) ∂ pˆ i ∂ ∂ pˆ i ∂ df (1) = −p + + . (4.159) + dt dp ∂t a ∂x i ∂t a ∂x i In Fourier space, we expand as 1 (r) = (2π )3
d3 k k eik·r ,
(4.160)
where the time dependence on the l.h.s. is implicit. Instead of the unit vector pˆ , we can use the direction cosine µ=
k · pˆ , k
where k = |k| is the magnitude of the wavevector k.
(4.161)
66
Basics of cosmological perturbation theory
It turns out convenient to integrate out the angular dependence of (k) by defining the multipoles 1 dµ 1 ≡ P (µ) (µ) , (4.162) (−i) −1 2 where P is the Legendre polynomial of order . The first polynomials are P0 (µ) = 1, P1 (µ) = µ, and P2 (µ) = (3µ2 − 1)/2. So = 0 defines the monopole 0 , which is simply the angular average of , while = 1 gives the dipole term. The next step is to evaluate the collision term C[f ] for the photons, determined mainly by the Compton scattering process e− + γ ↔ e− + γ . The interested reader may wish to consult the textbook [74], since its complete derivation requires several pages. The final result is however pretty simple: C[f ] = −p
∂f (0) ˆ + pˆ · v b ] , ne σT [ 0 − (p) ∂p
(4.163)
where ne is the electron density, σT is the Thomson cross section, and v b is the velocity of the electrons (here classified as usual in cosmology among the “baryons”). In the absence of v b , a strong coupling (i.e. the limit of large ne σT ) means that ≈ 0 , i.e. that the perturbation is fully described by its monopole term. Equating (4.159) and (4.163) we obtain ∂ pˆ i ∂ ∂ pˆ i ∂ ˆ + pˆ · v b ] . + + + = ne σT [ 0 − (p) ∂t a ∂x i ∂t a ∂x i
(4.164)
In Fourier space we have that ∂ /∂x j → ikj , and as usual we omit the subscript k for Fourier-transformed quantities. Moreover, since we always assume the fluid to be irrotational, v b is directed along k so that vbi = vb k i /k ,
(4.165)
where vb is the velocity modulus. Then Eq. (4.164) can be written as ∂ ikµ ∂ ikµ ˆ + µvb ] . + + + = ne σT [ 0 − (p) ∂t a ∂t a
(4.166)
In terms of the conformal time η, this equation reduces to ˆ + µvb ] , + ikµ + + ikµ = −τop [ 0 − (p)
where we have introduced the optical depth τop defined as η0 τop ≡ ne σT a dη˜ . η
(4.167)
(4.168)
4.9 Baryons, photons, and neutrinos
67
Note that a prime represents a derivative with respect to η (with η0 being the conformal time today). The optical depth is small when the free electron density ne is small, i.e. at late times. In the limit τop → 0, the photon fluctuations decouple from the baryons. In this limit, i.e. putting the r.h.s. of Eq. (4.167) to zero, the equation can be applied as well to massless neutrinos. On the contrary, when ne is large such that τop 1, i.e. before the recombination epoch, the photons and baryons are tightly coupled. Now we should redo the calculations for massive particles, in order to derive the Boltzmann equation for baryons. Instead of the momentum magnitude p, we introduce the energy E = p 2 + m2 . For non-relativistic matter, all terms of order v 2 = (p/E)2 or higher can be neglected. This allows us to write the Boltzmann equations for the distribution fm in terms of the first and second moments, i.e. of the number density and velocity nm ≡
d3 p fm , (2π )3
i vm ≡
1 nm
d3 p p pˆ i fm . (2π )3 E
(4.169)
Recall that in Section 4.3 we have derived the set of perfect fluid equations (4.48) and (4.51) for the pressureless matter component. For the baryons, however, there is again a collisional term on the r.h.s. of the Boltzmann equation. This term contributes only to the velocity equation (4.51), i.e. the first moment of the baryon distribution fb , since both Compton scattering (that couples electrons and photons) and Coulomb scattering (that couples electrons and protons) conserve the number of “baryons.” Therefore the velocity equation acquires an additional term coming from the scattering process. In Fourier space the full set of coupled photon–baryon equations reads ( 0 − + µvb ) , = − − ikµ( + ) − τop
(4.170)
δb = −ikvb − 3 ,
(4.171)
vb = −Hvb − ik +
τop
Rs
(3i 1 + vb ) ,
(4.172)
3 ρb . 4 ργ
(4.173)
where Rs ≡
j
Note that we have used the relation θb = ikj vb = ikvb to convert θb to vb in Eqs. (4.171) and (4.172). The cold dark matter does not have an electromagnetic interaction so that the perturbations are not affected by the collision term unlike the baryons. Hence the
68
Basics of cosmological perturbation theory
CDM perturbations δc and vc obey the following equations δc = −ikvc − 3 ,
(4.174)
vc = −Hvc − ik ,
(4.175)
which are of course identical to the matter equations we derived in Eqs. (4.86) and (4.87) (with = −). Now we can derive an equation for the monopole term 0 that will be useful later on. Multiplying P0 (µ) and P1 (µ) for Eq. (4.170) and integrating them in terms of µ in the range [−1, 1], we obtain 0 + k 1 = − ,
k i 1 − v b 1 − ( 0 + ) = τop 3 3
(4.176)
,
(4.177)
where we have neglected the term 2 (see problem 4.2). In the tight coupling regime 1 | | 1 | in Eq. (4.177). Then the second term, (k/3) 0 , (τop 1) one has |τop ∼ (kη/τop ) 0 . in Eq. (4.177) is the same order as the r.h.s., giving 1 ∼ k 0 /τop In the tight coupling limit (τop 1) this means that 1 0 for the modes around the Hubble radius (kη ∼ 1). This property generally holds for the multipole moments , i.e. +1 / ∼ kη/τop . Hence it is a good approximation to neglect the moments ( ≥ 2) compared to monopole and dipole moments in the tight coupling regime. )[vb + Hvb + ik], If we rewrite Eq. (4.172) in the form vb = −3i 1 + (Rs /τop then the term (Rs /τop )[vb + Hvb + ik] is suppressed relative to others by a factor of 1/τop , giving the relation vb −3i 1 at the lowest order. Substituting this relation into Eq. (4.172), we get Rs k vb −3i 1 − 3i 1 + H 1 − . (4.178) τop 3 Plugging Eq. (4.178) into Eq. (4.177) and taking the η derivative of Eq. (4.176) to eliminate the term , we find (see the problem 4.3) 0 +
Rs Rs k2 H 0 + k 2 cs2 0 = − − H − , 1 + Rs 3 1 + Rs
(4.179)
where cs2 ≡
δPγ 1 = δργ + δρb 3(1 + Rs )
(4.180)
is the effective sound speed squared of the coupled baryon–photon plasma. Recall that Rs is defined in Eq. (4.173). The last equality in Eq. (4.180) holds by using the relations δPγ = δργ /3 and δργ /ργ = (4/3)δρb /ρb (the second one comes from
4.9 Baryons, photons, and neutrinos
69
ργ ∝ a −4 ∝ T 4 and ρb ∝ a −3 ∝ T 3 ). In the absence of baryons the sound speed √ reduces to that for a relativistic fluid: cs = c/ 3, where we have recovered the speed of light c. When baryons are present, the sound speed decreases because the fluid becomes heavier. Equation (4.179) is very convenient because it shows how the direction-averaged temperature anisotropy 0 responds to gravity, i.e. to and . In Section 5.3 we shall discuss the solution of this equation to confront the predicted temperature anisotropies with CMB observations. The gravitational potentials and are determined by the combined action of all the matter fields, baryons (b), cold dark matter (c), photons (γ ), neutrinos (ν), and dark energy (DE). If we neglect the contribution of dark energy, the (00) component of the Einstein equation (4.57) reads (4.181) k 2 + 3H( − H) = 4π Ga 2 ρm δm + 4ρr r,0 , where ρm δm ≡ ρb δb + ρc δc , ρr r,i ≡ ργ i + ρν Ni ,
(4.182) (i = 0, 1, 2, . . .).
(4.183)
Here 0 and N0 are the monopoles of photons and massless neutrinos, respectively, which are related to the energy momentum tensors via (γ ) T00 = −ργ (1 + 4 0 ) and (ν) 0 T0 = −ρν (1 + 4N0 ). From Eqs. (4.58) and (4.181) it follows that 3H 2 2 (iρm vm + 4ρr r,1 ) , k = 4π Ga ρm δm + 4ρr r,0 + (4.184) k where we have used the fact that the velocity vr for radiation is related to the dipole r,1 via vr = −3i r,1 [74]. There is also another relation that relates the sum of the gravitational potentials and with the anisotropic stresses of photons and neutrinos: k 2 ( + ) = −32π Ga 2 ρr r,2 .
(4.185)
If the quadrupole r,2 is negligibly small, it follows that −. Equation (4.185) assumes a particularly simple form during the radiation-dominated era and for super-horizon scales. During the radiation-dominated era we can use the Friedmann equation (2.17) in flat space to rewrite it as k 2 ( + ) = −12H2 fν N2 ,
(4.186)
where fν =
ρν . ρν + ργ
(4.187)
70
Basics of cosmological perturbation theory
In fact, the dominant term in the quadrupole r,2 is the neutrinos since the photon– baryon plasma behaves almost as a perfect fluid. Note that fν ≈ ν ≈ constant during the radiation-dominated era but not afterward. For adiabatic initial conditions the large-scale fluctuations in radiation equal the fluctuations in the neutrino component since they are both relativistic. Therefore we can put r,0 = N0 . Now from Eq. (4.181), assuming k H and constant (i.e. discarding the first and second term on the l.h.s.) and negligible ρm , we obtain 3H2 = −16π Ga 2 ρr r,0 .
(4.188)
= −2 r,0 = −2N0 .
(4.189)
It then follows that
This replaces Eq. (4.71) when we cannot assume = −. For the neutrinos we can use Eqs. (4.176) and (4.177) with N0,1 instead of 0,1 and no coupling to baryons. However the quadrupole N2 that we neglected in Eq. (4.177) is no longer negligible. Multiplying Eq. (4.170) by P2 (µ) and integrating it with respect to µ, we obtain an equation for N2 . Neglecting N3 for k H and eliminating N1 , one finds 2k 2 ( + N0 − 2N2 ) . 15 Equation (4.186) can be written as N2 =
N2 = AH−2 ,
(4.190)
(4.191)
where A = −k 2 ( + )/(12fν ) is approximately constant since the potentials are constant on super-horizon scales. Differentiating Eq. (4.191) twice with respect to η (notice that H /H2 −1 = constant in this regime) we obtain N2 = 2A = −k 2 ( + )/(6fν ) .
(4.192)
Combining Eqs. (4.189), (4.190), (4.192) and taking the limit k H, it follows that 2 (4.193) = − 1 + fν . 5 This sets the initial condition for the relation between the potentials at early times. 4.10 The matter power spectrum The large-scale structure of the Universe started to grow after the epoch of the radiation–matter equality. Since non-relativistic matter has a negligible pressure relative to its energy density, the gravitational attraction becomes stronger than the
4.10 The matter power spectrum
71
pressure repulsion in the matter-dominated epoch. The perturbations of pressureless matter, especially the CDM perturbations, are responsible for the formation of galaxies. We can quantify the matter distribution in the Universe by measuring the correlation function or the power spectrum of the galaxies we observe in the sky. In order to derive the spectrum of matter perturbations today, we need to know the evolution of (k, t) from the early Universe to the present epoch. The cosmic inflation in the early Universe that is needed to solve the problems of flatness and horizon is believed to have generated the seeds of anisotropies through the quantum fluctuations of a scalar field (called “inflaton”) [70] . In the simplest version of the single-field inflation scenario the perturbations are “frozen” after the scale λp = (2π/k)a of perturbations leaves the Hubble radius (H −1 ) during the accelerated expansion (we are considering the perturbations whose wavelengths are smaller than the Hubble radius at the onset of inflation). The perturbations in the super-horizon regime (characterized by k H = aH ) evolve only adiabatically after the first Hubble radius crossing (k = H). After inflation the comoving Hubble radius, H−1 , begins to increase as long as the Universe undergoes a decelerated expansion (a ∝ t p with p < 1). During radiation and matter eras, the perturbations cross inside the Hubble radius again (k H). The epoch of this second Hubble radius crossing (k = H) depends upon the wavelength of perturbations. The crossing occurs earlier for smaller scale perturbations. During inflation nearly scale-invariant density perturbations are generated from the quantum fluctuation of a scalar field with a potential, which sets up initial conditions for the gravitational potential. Inflation generally predicts nearly scaleinvariant density perturbations, which are consistent with the CMB anisotropies observed by COBE and WMAP. The initial power spectrum of generated during inflation is given by 50π 2 k ns −1 2 (i) 2 P ≡ |(k, ai )| = δH , (4.194) 9k 3 H0 (conventionally there is no factor of volume in this definition of power spectrum) 2 where ns is the spectral index and δH represents the amplitude of the gravitational potential (see Refs. [70, 74, 94] for the derivation of Eq. (4.194)). The WMAP 2 5-year observations constrain these values to be ns = 0.960+0.014 −0.013 and δH 3.2 × 10−10 [15]. Note that ns = 1 corresponds to the scale-invariant spectrum with k 3 |(k, ai )|2 = constant.1 1
Although in mathematics “scale-invariant” means generally a power-law behavior, cosmologists use this term (i) to denote a flat spectrum k 0 . In the context of inflation the power spectrum is usually defined by Pˆ ≡ 3 2 2 k /(2π )|(k, ai )| , but we adopt the definition (4.194) to match with the standard notation used to evaluate the matter power spectrum.
72
Basics of cosmological perturbation theory
In order to obtain the gravitational potential today, we need to solve the equation for (k, t) from the beginning of the radiation-dominated epoch to the present. The evolution of the gravitational potential during the early cosmological epoch depends on the modes k. The wavenumber keq that characterizes the border of “large-scale” and “small-scale” modes corresponds to the one that entered the Hubble radius at the radiation–matter equality, i.e. keq = aeq H (aeq ). Using Eq. (2.78) together with 3 1/2 H (aeq )/H0 = [2(0) , we find m /aeq ] $ 2(0) m 2 −1 = 0.073 (0) . (4.195) keq = H0 m h Mpc aeq Let us first consider the large-scale modes with k keq . In Section 4.4 we have already seen that there is a solution with = constant for a single-fluid system. However it remains to see what happens during the transition from the radiation era to the matter era. In the following we shall neglect the contribution of baryons as well as the quadrupole r,2 . Under the super-horizon approximation (k H), we obtain the following approximate equations from Eqs. (4.181), (4.174), and (4.176): 3H( + H) = 4π Ga 2 (ρc δc + 4ρr r,0 ) ,
(4.196)
δc = −3 ,
(4.197)
r,0 = − .
(4.198)
In Eqs. (4.196) and (4.198), r,0 includes the contributions of both radiation and neutrinos. From Eqs. (4.197) and (4.198) it follows that δc = 3 r,0 + C, where C is an integration constant. As we have already mentioned in the previous section, the radiation perturbations δr = δρr /ρr are related to the matter perturbation δm = δρm /ρm via δr = (4/3)δm by noting that ρr ∝ T 4 and ρm ∝ T 3 . This translates into δm = 3 r,0 , which is called the adiabatic condition (applied to both dark matter and baryons). This shows that C = 0 for CDM as long as the adiabatic initial condition is respected. Plugging r,0 = δc /3 into Eq. (4.196) and introducing the variable y ≡ a/aeq = ρc /ρr , we find y
d 3y + 4 += δc , dy 6(y + 1)
(4.199)
where we have used the relations dy/dη = Hy and 3H 2 = 8π Gρc (1 + 1/y). Taking the y-derivative of Eq. (4.199) and using Eq. (4.197), we obtain d2 21y 2 + 54y + 32 d 1 + + = 0. 2 dy 2y(y + 1)(3y + 4) dy y(y + 1)(3y + 4)
(4.200)
4.10 The matter power spectrum
This has the following analytic solution [90] √ y+1 9y 3 + 2y 2 − 8y − 16 + c , (y) = c1 2 y3 y3
73
(4.201)
where the integration constants c1 and c2 are determined by imposing the initial conditions i = (0) and (d/dy)i = 0. This gives c1 = 16c2 = (8/5)(0), so that the solution is √ 9y 3 + 2y 2 − 8y − 16 + 16 y + 1 (y) = (0) . (4.202) 10y 3 In the limit that y = a/aeq → ∞, the gravitational potential approaches → (9/10)(0). Hence, for super-horizon perturbations, the gravitational potential decreases by 10% during the transition from the radiation era to the matter era. Let us next consider the evolution of during the deep radiation-dominated epoch. Since cs2 1/3, H −H2 , and H 1/η in this regime, Eq. (4.66) gives 4 k2 + + = 0 . η 3
(4.203)
The solution of this equation satisfying the initial conditions = I and d/dη = 0 at η = 0 is √ √ √ sin(kη/ 3) − (kη/ 3) cos(kη/ 3) . (4.204) (k, η) = 3I √ (kη/ 3)3 For the modes outside the Hubble radius (k H, i.e. kη 1) we have (k, η) ! 2 I 1 − (kη) /10 , which means that the gravitational potential is nearly constant, as we know already. The modes with k keq crossed inside the Hubble radius (kη > 1) before the radiation–matter equality. These small-scale perturbations started to decay after the Hubble radius crossing. Asymptotically (kη 1) the gravitational potential decreases as 1/(kη)2 with oscillations (during the radiation era). For larger k this decay started earlier, which implies that the amplitude of the resulting gravitational potential is suppressed for perturbations on smaller scales. After the Universe enters the matter-dominated epoch, the amplitude of approaches a constant value. Thus we have shown that the evolution of the gravitational potential depends on the scales of perturbations. In order to describe its evolution for each wavenumber k during the transfer epoch (from the radiation era to the epoch at a = aT ), we introduce the transfer function T (k) ≡
(k, aT ) , LS (k, aT )
(4.205)
74
Basics of cosmological perturbation theory
where LS (k, aT ) is the large-scale solution decreased by an amount 9/10 compared to the primordial value (k, ai ) generated during inflation: 9 (4.206) LS (k, aT ) = (k, ai ) . 10 The typical value of aT is around aT ∼ 0.03. Note that for a > aT the evolution of becomes independent of k (i.e. = constant during the matter era). In general, the transfer function has to be derived numerically by integrating the equations for each k mode. A popular fit has been given by Bardeen, Bond, Kaiser, and Szalay (BBKS) [100]: T (x) =
!−1/4 ln(1 + 0.171x) 1 + 0.284x + (1.18x)2 + (0.399x)3 + (0.490x)4 , 0.171x (4.207)
where x ≡ k/keq . On large scales characterized by the condition x = k/keq 1 the BBKS transfer function reduces to T (x) 1, which means that (k, aT ) = (9/10)(k, ai ) as expected. On small scales with x = k/keq 1 the transfer function has a k-dependence T (k) ∝ (ln k)/k 2 , which means that the gravitational potential (k, aT ) is suppressed for increasing k. Note that the logarithmic correction comes from matching two solutions of dark matter perturbations during the radiation and matter eras. When a > aT the gravitational potential remains constant during the matter era at least in standard General Relativity, but after the Universe has entered the epoch of cosmic acceleration is expected to vary. In order to quantify this, we introduce the growth function D(a): (a) D(a) = (aT ) a
(a > aT ) .
(4.208)
In Fig. 4.1 we illustrate the evolution of the quantity D(a)/a in both CDM and CDM models. In the CDM model the gravitational potential is constant for a > aT and hence D(a) = a. In the CDM model, the constant evolution of (a) is followed by its decay around the end of the matter era. This is associated with the decreases of both m and the growth rate of δm . This variation of leads to the integrated Sachs–Wolfe (ISW) effect in the CMB temperature anisotropies, as we will see in more detail in Sections 4.11 and 5.3. Combining Eqs. (4.205), (4.206), and (4.208), we find that the gravitational potential today is given by 9 (4.209) (k, a0 ) = (k, ai )T (k)D(a0 ) , 10 where we have used a0 = 1. After the Universe enters the matter era, the radiation perturbation r,0 is negligible relative to the matter perturbation δm . For the CDM model or the models
4.10 The matter power spectrum
75
1.2
Λ CDM CDM
D(a) / a
1.1
1.0
0.90
0.80
0.70 0
5
10
15
z Figure 4.1 The evolution of the quantity D(a)/a versus the redshift z for the (0) CDM model with (0) m = 0.28 and the CDM model with m = 1.0. Since the gravitational potential does not vary in time during the matter-dominated era, the quantity D(a)/a remains constant for the CDM model. However, in the CDM model, the decrease of the gravitational potential at late times leads to the variation of D(a)/a.
in which dark energy does not cluster, we can also ignore the dark energy perturbation compared to the matter perturbation. Under the sub-horizon approximation (k H), Eq. (4.181) then reduces to k 2 = 4π Ga 2 ρm δm .
(4.210)
2 (0) Using the relations ρm = ρm(0) /a 3 and (0) m = 8π Gρm /(3H0 ), Eq. (4.210) can be expressed as
δm (k, a) =
2k 2 a 2 3(0) m H0
(k, a) .
(4.211)
Using Eqs. (4.194), (4.209), and (4.211), the power spectrum of matter perturbations at the present epoch is ns k 2π 2 δ 2 T 2 (k)D 2 (a0 )H0−3 . (4.212) Pδm ≡ |δm (k, a0 )|2 = (0) H 2 H 0 (m ) On large scales characterized by x = k/keq 1 the matter power spectrum has a scale-dependence Pδm ∝ k ns , which grows for increasing k. On small scales (x = k/keq 1) one has Pδm ∝ k ns −4 (ln k)2 , which decreases for increasing k. Therefore there is a peak for the matter power spectrum, whose wavenumber is
76
Basics of cosmological perturbation theory
determined by keq given in Eq. (4.195). As we will see in Section 5.5, the matter power spectrum is important in measuring dark energy properties.
4.11 Perturbed photon propagation So far we have studied the evolution of matter density perturbations. However, what we really observe in cosmology is radiation emitted from sources. It is then time to face the question of the propagation of radiation in a perturbed Universe. That is, what happens to photons (their energy, their path) when they propagate through a perturbed space time? The general answer is provided by the two equations of light propagation (we do not consider polarization here). If k µ = dx µ /dλs is the photon momentum, then the null condition and the geodesic equation read k µ kµ = 0 ,
(4.213)
dk µ µ + αβ k α k β = 0 , dλs
(4.214)
where λs is an affine parameter that plays no explicit role because we can always use the µ = 0 geodesic equation to convert λs to the conformal time η. Solving these equations in the perturbed metric (4.13) gives the general equations of photon propagation. The solution will give the variation in the photon’s frequency and path due to the inhomogeneous metric. By observing the inhomogeneities in frequency and in the angular direction, we can constrain the perturbations that the photon went through during its long cosmic ride. We will see in Chapter 14 that these effects, in particular the ISW effect and the weak lensing, are very useful cosmological probes. Let us start by splitting the momentum vector k µ = dx µ /dλs into a background and a perturbed value k µ = kˆ µ + δk µ .
(4.215)
We consider a photon that, in the unperturbed metric, propagates along the direction r, so that the background propagation equation in flat space is dη = dr. In this section the perturbed quantities are in real space, not in Fourier space, since we need to deal with physical trajectories. Also we use the explicit derivation symbol ∂/∂η instead of the prime because we want to distinguish carefully between partial and total derivatives. At background level, the geodesic equation (index 0) gives simply dkˆ 0 = −2H(kˆ 0 )2 , dλs
(4.216)
4.11 Perturbed photon propagation
77
0 where we have used 00 = H for the flat FLRW metric with the conformal time η. Equation (4.216) is integrated to give
dη kˆ 0 = ∝ a −2 . dλs
(4.217)
This is all we need to convert λs to η. Since the photon frequency is ν ≡ dt/dλs = adη/dλs , we see that the frequency redshifts in proportion to 1/a, as it should. We also need the perturbed null-geodesic condition δk µ kˆµ + kˆ µ δkµ = 0 .
(4.218)
For radial trajectories the null path condition ds 2 = 0 gives (1 + )dη = (1 + )dr ,
(4.219)
dr dη = (1 + − ) , dλs dλs
(4.220)
kˆ 0 + δk 0 = (1 + − )(kˆ r + δk r ) .
(4.221)
or
which gives
We obtain kˆ 0 = kˆ r from the zero-th order equation. The first-order terms give δk 0 = δk r + kˆ 0 ( − ) .
(4.222)
Now we may be tempted to derive the first-order perturbed µ = 0 equation (4.214) straightforwardly as dδk 0 /dλs = −4Hkˆ 0 δk 0 − (kˆ 0 )2 (∂/∂η + ∂/∂η + 2,r ). However, that would be wrong, since the affine parameter λs needs to be perturbed as well. We cannot take it to be the same as for the zeroth-order case. The way to proceed is then to derive the full expression up to first-order: ∂ ∂ dk 0 0 2 0 0 0 2 (4.223) + + 2,r , = −2H[(kˆ ) + 2kˆ δk ] − (kˆ ) dλs ∂η ∂η where we used Eq. (4.222) and where k 0 = kˆ 0 + δk 0 , so there is only one affine parameter for the full momentum k 0 . The situation is simplified if we “complete” the expression (kˆ 0 )2 + 2kˆ 0 δk 0 with an extra second-order term (δk 0 )2 to obtain (k 0 )2 and similarly substitute (k 0 )2 for (kˆ 0 )2 . This is allowed because we are adding second-order terms that will be automatically discarded later on. Then we obtain dk 0 ∂ 0 2 0 2 ∂ (4.224) = −2H(k ) − (k ) + + 2,r . dλs ∂η ∂η
78
Basics of cosmological perturbation theory
Dividing Eq. (4.224) with respect to k 0 = dη/dλs , we can rewrite it in the form 1 d(a 2 k 0 ) ∂ ∂ (4.225) =− + + 2,r . a 2 k 0 dη ∂η ∂η Now we split the momentum k 0 to be k 0 = kˆ 0 + δk 0 . Using the fact that Eq. (4.216) is equivalent to d(a 2 kˆ 0 )/dη = 0, it follows that 1 d(a 2 k 0 ) d(δk 0 /k 0 ) → , a 2 k 0 dη dη
(4.226)
where, at first-order, we can put equivalently k 0 or kˆ 0 at the denominator. We then obtain the following equation d(δk 0 /k 0 ) ∂ ∂ (4.227) =− + + 2,r . dη ∂η ∂η For the spatial equations we need to derive only the two directions x 1 and x 2 orthogonal to the propagation direction r. For i = 1, 2, using the geodesic condition dη2 − dr 2 = 0, we find from Eq. (4.214) dη dx i dη 2 d2 x i + 2H = (,i − ,i ) . (4.228) dλ2s dλs dλs dλs Here we do not need to worry about the unperturbed λs because there is no unperturbed propagation along the directions x 1 and x 2 . We write this equation explicitly in terms of the perturbed coordinates x i (λs ) of the light ray instead of k i since this will be useful in the weak lensing observation. Notice that in several derivations at this stage one encounters additional terms proportional to k i ,i which are then eliminated by invoking the so-called thin-lens approximation. Since we are considering first-order quantities, these terms are automatically canceled from the beginning. Now the problem splits neatly in two: (i) the solution to Eq. (4.227) leads to the discussion of the Sachs–Wolfe effect, i.e. the change of a photon’s redshift in passing through a gravitational potential, (ii) the solution to Eq. (4.228) leads to weak lensing, i.e. the deviation of a light ray passing through the same.
4.11.1 The Sachs–Wolfe effect Let us begin with Eq. (4.227). First we notice the following relation ˆk 0 ∂ + ,r = ∂ dη + ∂ dr = d → ∂ + ,r = d , ∂η ∂η dλs ∂r dλs dλs ∂η dη (4.229)
4.11 Perturbed photon propagation
79
where we have used kˆ 0 = dη/dλs and also dη = dr along the unperturbed light ray. We then find that Eq. (4.227) can be written as d d(δk 0 /k 0 ) = −2 − dη dη
∂ ∂ − ∂η ∂η
,
(4.230)
which is integrated to give # O δk 0 ##O ∂ ∂ O − dη , = −2| − E k 0 #E ∂η ∂η E
(4.231)
where “O” and “E” represent the instants of observation and emission, respectively, O and |O E ≡ O − E . Notice that the last term does not integrate to ( − )E since the potentials must be calculated along the light-ray null path r(η) and the integral contains the partial, not total, derivative with respect to η. This is our first result. The frequency shifts by an amount that depends on the difference between the potential at emission and at observation and on the lineof-sight integral of − . The first effect is referred to as the Sachs–Wolfe effect (although often another term gets counted as Sachs–Wolfe, see below) and the second as the integrated Sachs–Wolfe effect. The most important application of this result is the calculation of the CMB temperature anisotropies. The temperature of a black body distribution of photons is proportional to their average frequency νˆ . More exactly, for photons propagating with four-momentum k µ and emitted by a body moving with respect to us with velocity uµ , the temperature we measure is proportional to −k µ uµ = νˆ . The same applies if uµ is our own velocity with respect to the source. Therefore if a photon is emitted at E and observed at O, the emission temperature TE and observed temperature TO are related with each other via TO (k µ uµ )O = µ . TE (k uµ )E
(4.232)
The observed temperature fluctuation is # # # # δT ## δT ## δ(k µ uµ ) ## δ(k µ uµ ) ## = + µ − µ . T #O T #E k uµ #O k uµ #E
(4.233)
As usual we assume a radial propagation along the coordinate r, i.e. dη = dr. Recall that the four-velocity is given by uµ = [−a(1 + ), avi ] for the perturbed metric (4.13). Notice that at zero-th order we have k µ uµ = k 0 u0 . Let us consider
80
Basics of cosmological perturbation theory
the second term on the r.h.s. of Eq. (4.233). The non-zero first-order elements are δu0 k 0 δu0 = =, µ k uµ u0
(4.234)
k i δui ki k i δui = = − vi = ei vi , k µ uµ k 0 u0 k0
(4.235)
where ei ≡ −k i /k 0 is the unit direction vector, and the term (δk 0 )u0 /(k µ uµ ) = δk 0 /k 0 is given in Eq. (4.231). We can now put the pieces together. We find # # O ∂ ∂ δT ## δT ## i O O dη . (4.236) − = + e v | − | + i E E T #O T #E ∂η ∂η E The term |O E = O − E can be further simplified since the temperature distortion induced by the local gravitational potential O is unobservable (it affects equally the whole CMB). Then we are left with four other terms in Eq. (4.236), which in turn are called the intrinsic temperature fluctuation (δT /T |E ), the Doppler Sachs–Wolfe (SW) effect (E ), and the integrated Sachs–Wolfe shift (ei vi |O E ), the O (ISW) effect ( E (∂/∂η − ∂/∂η)dη), respectively. When we discuss photons coming from the last scattering surface of the CMB, the intrinsic temperature fluctuation is also counted as the SW effect (the separation between these effects is in fact gauge dependent). As we have mentioned in Section 4.10 the adiabatic initial condition of the baryon–photon fluid system in the tight coupling regime corresponds to δm = 3δT /T |E , where δm is the matter perturbation. From Eq. (4.71) the matter perturbation satisfies the relation δm = −2 on scales larger than the Hubble radius as long as the matter dominates the overall fluctuations and the anisotropic stress is neglected such that = −. So finally one arrives at the (adiabatic, large-scale, shear-free) familiar result # δT ## 2 1 = − E + E = E , (4.237) # T SW 3 3 where the subscript E refers to the epoch of last scattering. Since the dark energy density in most models is negligible at early times, E is hardly affected by it. In contrast, the last term in Eq. (4.236), the ISW term, is a line-of-sight term and does therefore contain potentially much information on the recent Universe. Moreover, we have already seen that the gravitational potential is constant for a matter-dominated Universe and therefore yields no ISW signal. This shows that the ISW effect is a direct diagnostic of something which is not ordinary pressureless matter, just as we are looking for. The constraint on dark energy models through the ISW effect will be further discussed in Section 14.1.1.
4.11 Perturbed photon propagation
81
4.11.2 Weak lensing Let us now tackle the spatial part of the geodesic equations. Using Eqs. (4.228) and (4.217), we obtain the propagation equation for i = 1, 2: d2 x i = ψ,i , dr 2 where ψ is the lensing potential defined by ψ ≡−.
(4.238)
(4.239)
In standard General Relativity with ordinary matter, ψ = −2. Since the displacement vector x = (x 1 , x 2 ) is small, we can put x i = rθ i and hence Eq. (4.238) is written as d2 (rθ i ) = ψ,i . (4.240) dr 2 If the light ray reaches the observer located at r = 0 through the direction θ0i = (θ01 , θ02 ) the integration of Eq. (4.240) twice leads to (here and in the rest of this section we use primes to denote dummy integration variables, not derivatives) 1 r r i i θ = θ0 + dr dr ψ,i (r θ01 , r θ02 , r ) , (4.241) r 0 0 where the integration constant is to be equal to the observed angle θ0i so that the angle remains the same in the absence of metric perturbations (ψ = 0). The region of the integral with respect to r and r is restricted to be in the region 0 < r < r, 0 < r < r , or in other words, r < r < r, 0 < r < r. Carrying out the r integral of Eq. (4.241) in the latter region, it follows that r r i i ψ,i (r θ01 , r θ02 , r ) . θ = θ0 + dr 1 − (4.242) r 0 Two light rays separated by a small interval x will obey the equation r r j r ψ,ij (r θ01 , r θ02 , r ) , dr 1 − (4.243) θ i = θ0i + θ0 r 0 where the term r ψ,ij arises by taking the variation of ψ,i with respect to θ0 (j = 1, 2). Therefore, if the separation θ i is taken on the source plane at r = rs , we have an equation that connects the source plane with the observation plane at r = 0. The entire phenomenon can be described by the symmetric transformation matrix (see Fig. 4.2) j
Aij ≡
∂θsi j
∂θ0
= δij + Dij ,
(4.244)
82
Basics of cosmological perturbation theory
qS Source plane
q0 Image or Lensing plane
Observer
Figure 4.2 The geometry of the weak lensing.
where Dij =
rs
dr 0
r 1− rs
r ψ,ij =
−κwl − γ1 −γ2
−γ2 −κwl + γ1
is the distortion tensor. The parameter r 1 rs r (ψ,11 +ψ,22 ) , dr 1 − κwl = − 2 0 rs
,
(4.245)
(4.246)
is called the convergence and describes the magnification of the source image. The quantities γ1 , γ2 are the two components of the shear field r 1 rs r (ψ,11 −ψ,22 ) , dr 1 − (4.247) γ1 = − 2 0 rs rs r γ2 = − r ψ,12 , dr 1 − (4.248) rs 0 and describe the distortion of the source image. Although in principle both magnification and shape distortion could be used as cosmological tools, the noise in the former (i.e. the large intrinsic variation of galaxy luminosities) has not allowed so far a practical use of the magnification in cosmology. In Section 14.4 we will study how to employ the distortion as a test of cosmology.
4.12 Problems
83
4.12 Problems 4.1 Let the distribution function of the density contrast δx be given by the Gaussian 1 δx2 (4.249) exp − 2 . Px (δx ) = (2π σ 2 )1/2 2σ Suppose that the fluctuations δx , δy at any two given points x, y separated by r are distributed as a two-dimensional Gaussian field with the probability distribution 1 1 −1 δ , (4.250) Pxy (δx , δy ) = exp − C δ i j ij 2 1/2 2 2π(σ 4 − ξ12 ) where δi = (δx , δy ) and the covariance matrix is 2 σ ξ12 (r) . Cij ≡ δi δj Pxy (δx , δy )dδx dδy = σ2 ξ12 (r)
(4.251)
Let us now cut the field so that we consider only fluctuations above a given threshold. r (1) Find the general expression of the probability of having δ > νσ given that x δy > νσ at a distance |x − y| = r, where ν is a positive parameter that defines a threshold in units of σ . r (2) Suppose that we divide a galaxy distribution into many small equal-volume regions above threshold (δ > νσ ) and below threshold. Find the expected value of the fraction of regions above threshold within a distance r of a region above threshold, divide this for the fraction of regions above threshold and interpret this as the correlation function 1 + ξ>b of regions above threshold. r (3) Find the limit of ξ 1 and ν 1 and interpret b = ν/σ as the bias of regions 12 above threshold with respect to the general field. (From Ref. [101].) Useful relations are: ∞ x2 ν 1 1 erfc √ e− 2σ 2 dx , (4.252) ≡ √ 2 2π σ νσ 2 e−x erfc(x) √ , x π 2
for
x 1,
2 2 erfc[x(1 + ε)] erfc(x) − √ εxe−x , π
(4.253) for
ε 1.
(4.254)
4.2 Derive Eqs. (4.176) and (4.177) from Eq. (4.170) by neglecting the contribution of the quadrupole moment 2 . 4.3 Confirm that Eq. (4.179) follows from Eqs. (4.176)–(4.178).
5 Observational evidence of dark energy
The existence of dark energy is supported by a number of observations. This includes (i) the age of the Universe compared to oldest stars, (ii) supernovae observations, (iii) Cosmic Microwave Background (CMB), (iv) baryon acoustic oscillations (BAO), and (v) large-scale structure (LSS). Even before 1998 it was known that in a CDM Universe the cosmic age can be smaller than the age of the oldest stars. Dark energy can account for this discrepancy because its presence can make the cosmic age longer. The first strong evidence for the acceleration of the Universe today came however by measuring the luminosity distance of the type Ia supernovae (SN Ia). The CMB observations are also consistent with the presence of dark energy, although the constraint coming from the CMB alone is not so strong. The measurements of BAO have provided another independent test for the existence of dark energy. The power spectrum of matter distributions also favors a Universe with dark energy rather than the CDM Universe. In the following we shall discuss this observational evidence for dark energy. The statistical method used to constrain cosmological parameters will be discussed in Chapter 13. More details on present and future observational aspects to detect dark energy will be presented in Chapter 14.
5.1 The age of the Universe As we already mentioned, the inverse of the Hubble constant H0 is a rough measure of the age t0 of the Universe. Here we shall compute t0 more precisely and compare it with the age of the oldest stars. For simplicity we assume that the equation of (0) (1 + z)3(1+wDE ) state of dark energy is a constant, in which case we have ρDE = ρDE from Eq. (2.81). Taking into account radiation, non-relativistic matter, and dark energy as components of the Universe, Eq. (2.84) gives the Hubble parameter H (z)
84
5.1 The age of the Universe
85
normalized by H0 : 1/2 (0) (0) 4 (0) 3 3(1+wDE ) 2 (1 + z) + (1 + z) + (1 + z) + (1 + z) . E(z) = (0) r m K DE (5.1) On using the relation dt = −dz/[(1 + z)H ], the age of the Universe is expressed as ∞ dz −1 . (5.2) t 0 = H0 E(z) (1 + z) 0 The integral (5.2) is dominated by the terms at low redshifts. Since (0) r is of the order of 10−5 –10−4 from Eq. (2.47), radiation becomes important only for high redshifts (z 1000). Hence it is a good approximation to neglect the contribution from radiation when we evaluate Eq. (5.2). Let us consider the case of the cosmological constant (wDE = −1). Then the age of the Universe is given by ∞ dx t0 = H0−1 (5.3) 1/2 , (0) 3 (0) (0) 2 1 x m x + DE + K x (0) (0) where x ≡ 1 + z and (0) m + DE + K = 1. (0) For the flat Universe (K = 0), Eq. (5.3) is integrated to give (0) −1 1 − 1 + m H , ln t0 = 0 (0) (0) 3 1 − m 1 − 1 − m
(5.4)
(0) (0) where we have used the relation (0) m + DE = 1. In the limit DE → 0 we have
2 (5.5) t0 = H0−1 . 3 On using the value (2.36) together with h = 0.72 ± 0.08, the age of the Universe in the absence of the cosmological constant is in the range 8.2 Gyr < t0 < 10.2 Gyr. Carretta et al. [102] estimated the age of globular clusters in the Milky Way to be 12.9 ± 2.9 Gyr, whereas Jimenez et al. [103] obtained the value 13.5 ± 2 Gyr. Hansen et al. [104] constrained the age of the globular cluster M4 to be 12.7 ± 0.7 Gyr by using the method of the white dwarf cooling sequence. In most cases the ages of globular clusters are larger than 11 Gyr. Then the cosmic age estimated by Eq. (5.5) is inconsistent with the ages of the oldest globular clusters. This problem can be circumvented by taking into account the cosmological constant (or dark energy with an equation of state wDE close to −1). Equation (0) (5.4) shows that t0 gets larger for decreasing (0) m . In the limit m → 0 we have t0 → ∞. In Fig. 5.1 we plot the cosmic age (5.4) versus (0) m together with the
86
Observational evidence of dark energy
Flat universe with Λ
1.4
H0 t 0
1.2
The WMAP bound 1.0
The allowed region from oldest stars
0.80
Open Universe without Λ 0.60 0
0.2
0.4
0.6
0.8
1
Ωm(0) Figure 5.1 The cosmic age t0 in the unit of H0−1 versus (0) m . The thin-solid curve describes a flat Universe in the presence of the cosmological constant with the (0) relation (0) m + DE = 1. The dashed curve corresponds to an open Universe without the cosmological constant. The thick-solid line is a minimum age allowed from the ages of oldest globular clusters (> 11 Gyr). We also show the bound coming from the WMAP 5-year data with h = 0.70. The flat Universe with the cosmological constant is consistent with the WMAP bound for 0.271 < (0) m < 0.289.
boundary allowed from the oldest stellar ages. In order to satisfy the condition t0 > 11 Gyr we require that 0 < (0) m < 0.55. The WMAP 5-year constraint on the cosmic age (assuming the CDM model) is given by t0 = 13.73 ± 0.12 Gyr [15]. Under this bound we find that the density parameter of non-relativistic matter is constrained to be 0.245 < (0) m < 0.261 for h = 0.72. Of course this bound changes if different values of h are chosen. When h = 0.70 the constraint becomes 0.271 < (0) m < 0.289 (see Fig. 5.1). In the open Universe ((0) K > 0) it is also possible to make the cosmic age larger −1 than (2/3)H0 even in the absence of dark energy. Setting (0) DE = 0 in Eq. (5.3), we obtain (0) −1 (0) 1 − 1 − m H0 1 + m , ln (5.6) t0 = (0) (0) (0) 1 − m 2 1 − m 1 + 1 − m (0) (0) where (0) m + K = 1. In the limit m → 1 we recover the value (5.5) in the −1 flat Universe. Meanwhile, in the limit (0) m → 0, we have t0 → H0 . The cosmic
5.2 Supernova observations
87
age in the open Universe does not become so large compared to the case of the flat Universe with the cosmological constant. Since the curvature |(0) K | has been constrained to be much smaller than unity from the WMAP measurements [15], it is not possible to satisfy the condition t0 > 11 Gyr for h = 0.72 ± 0.08 in the open Universe without dark energy. The above discussion shows that the existence of dark energy is crucially important to solve the cosmic age problem. 5.2 Supernova observations In 1998 Riess et al. [High-redshift Supernova Search Team (HSST)] [1] and Perlmutter et al. [Supernova Cosmology Project (SCP)] [2] independently reported the late-time cosmic acceleration by observing distant supernovae of type Ia (SN Ia). Up to 1998 Riess et al. had discovered 16 high-redshift SN Ia together with 34 nearby supernovae, while Perlmutter et al. had found 42 supernovae in the redshift range z = 0.18–0.83. The explosion of supernovae is extremely luminous and causes a burst of radiation. The supernovae can be classified according to the absorption lines of chemical elements. If the spectrum of a supernova includes a spectral line of hydrogen, it is classified Type II. Otherwise it is called Type I. If a supernova contains an absorption line of singly ionized silicon, it is further classified Type Ia (note that Type Ib contains a line of helium, whereas Type Ic lacks the lines of both silicon and helium). The explosion of Type Ia occurs when the mass of a white dwarf in a binary system exceeds the Chandrasekhar limit [105] by absorbing gas from another star. Since the absolute luminosity of Type Ia is almost constant at the peak of brightness, the distance to a SN Ia can be determined by measuring its observed (apparent) luminosity. Thus the SN Ia is a kind of “standard candle” by which luminosity distance can be measured observationally. In reality things are more complicated than this simple view. The intrinsic spread in absolute magnitudes is actually too large to produce stringent cosmological constraints. However, at the end of the 1990s, a high-quality sample of “local” (i.e. z 1) supernovae allowed the absolute magnitude to be correlated with the width of the light curve [106]: brighter supernovae have a broader light curve. By measuring at the same time the apparent magnitude and the light curve it is possible therefore to predict the absolute magnitude. Although in the following we refer to a universal SN Ia absolute magnitude, we always mean the magnitude corrected for the light curve width. The apparent magnitude, m, is often used as a measure of brightness of stars observed on Earth. Let us consider two stellar objects whose apparent fluxes are given by F 1 and F2 . The apparent magnitudes of those stars (m1 and m2 ) are related
88
Observational evidence of dark energy
to the fluxes according to
F1 5 . m1 − m2 = − log10 2 F2
(5.7)
This implies that a star with m1 = 1 is about 100 times brighter than one with m2 = 6. From the definition (5.7) the apparent magnitude is smaller for brighter objects. Choosing an appropriate reference flux, for Sun and Moon we have m = −26.7 and m = −12.6, respectively. We define the absolute magnitude M of an object in terms of an apparent magnitude m and a luminosity distance dL : dL . (5.8) m − M = 5 log10 10 pc If the distance is expressed in Megaparsec then the relation can be written as m − M = 5 log10 dL + 25 .
(5.9)
In other words the absolute magnitude corresponds to the apparent magnitude the object would have if it were located at the luminosity distance dL = 10 pc from the observer. An additional correction, denoted as K-correction, is due to the fact that as the redshift increases we observe different parts of the source spectrum: we always assume that this correction has been already included in the estimation of m. The absolute magnitude of SN Ia is known to be around M = −19 at the peak of brightness. If we consider two SN Ia whose apparent magnitudes are m1 and m2 with luminosity distances dL1 and dL2 , respectively, we obtain the following relation from Eq. (5.9): dL1 . (5.10) m1 − m2 = 5 log10 dL2 Since the observed flux F is proportional to dL −2 from Eq. (2.63), we find that Eq. (5.7) is consistent with Eq. (5.10). Since the (corrected) peak absolute magnitude M is the same for any SN Ia under the assumption of standard candles, the luminosity distance dL (z) is obtained from Eq. (5.9) by observing the apparent magnitude m. The redshift z of the corresponding SN Ia can be found by measuring the wavelength λ of light [see Eq. (2.28)]. The observations of many SN Ia provide the dependence of the luminosity distance dL in terms of z. Comparing observational data with the theoretical distance (2.68), it is possible to know the expansion history of the Universe for the redshift z O(1). Let us consider the case in which the Universe is dominated by a non-relativistic fluid and dark energy with an equation of state wDE . In this case the Hubble parameter H (z) is given by Eq. (2.84) with (0) r 0. Using the expansion (2.69)
5.2 Supernova observations
89
8.0 Flat with ΩD E = 0 Open with Ω K = 0.0085, ΩD E = 0
7.0
Flat with ΩD E = 0.7
H0 d L / c
6.0
5.0
4.0
3.0
2.0
1.0
0.0 0
0.5
1
1.5
2
2.5
3
3.5
4
z Figure 5.2 The luminosity distance dL versus the redshift z for three cases: (a) a flat Universe without dark energy, (b) an open Universe ((0) K = 0.0085) without dark energy, and (c) a flat Universe with the cosmological constant ((0) DE = 0.7 and wDE = −1). The presence of dark energy leads to a larger luminosity distance relative to the case without it. In the open Universe the luminosity distance also gets larger than that in the flat Universe.
around z = 0 we find that the luminosity distance, in the region z 1, is given by
1 c (0) (0) 2 3 z+ (5.11) dL (z) = 1 − 3wDE DE + K z + O(z ) . H0 4 In the flat Universe without dark energy we have dL (z) = (c/H0 )[z + z2 /4 + O(z3 )]. In the presence of dark energy (wDE < 0 and (0) DE > 0) the luminosity distance gets larger (see Fig. 5.2). Especially for smaller (negative) wDE and for larger (0) DE this tendency becomes more significant. In an open Universe (K < 0) the effect of the cosmic curvature also leads to a larger luminosity distance compared to the flat Universe. Since the curvature of the Universe is constrained to be 2 2 close to the flat one (−0.0175 < (0) K = −K/(a0 H0 ) < 0.0085 [15]) from WMAP 5-year data, it is difficult to give rise to a significant difference relative to the flat Universe without dark energy. This property can be seen in Fig. 5.2, which shows that the difference is small in the region z < 1.5. In 1998 Riess et al. [1] and Perlmutter et al. [2] released observational data of the apparent luminosity of high-redshift Type Ia supernovae (0.2 z 0.8). The data of low-redshift regions (z < 0.1) reported previously was also used in their analysis.
90
Observational evidence of dark energy
Let us pick up a few examples of data to understand how the luminosity distance is known observationally. First, consider two data of the apparent magnitudes in the low-redshift region of SN Ia: (i) 1990O: m = 16.26 (z = 0.03) and (ii) 1992bg: m = 16.66 (z = 0.036). Since the luminosity distance in the region z 1 is well approximated by dL cz/H0 from Eq. (5.11), the absolute magnitude M is known from Eq. (5.9). We take the value h = 0.7 for the Hubble constant given in Eq. (2.33). We then obtain M = −19.29 and M = −19.28 for 1990O and 1992bg, respectively. This shows that the absolute luminosity of SN Ia is nearly constant (M −19), as we already mentioned. Let us next use the high-redshift data reported by Perlmutter et al. [2]. Consider the two SN Ia data of the apparent magnitudes: (a) 1997R: m = 23.83 (z = 0.657), (b) 1995ck: m = 23.57 (z = 0.656). Employing the value M = −19.15 for the absolute magnitude, we find from Eq. (5.9) that the luminosity distance is given by H0 dL /c = 0.920 for 1997R and H0 dL /c = 0.817 for 1995ck. Notice that the approximation, dL cz/H0 , is no longer valid in the high-redshift regime. Let us consider a flat Universe with a dark energy equation of state wDE = −1 (i.e. (0) 1/2 3 in this case, the the cosmological constant). Since E(z) = [(0) m (1 + z) + DE ] luminosity distance (2.68) reads d˜z c(1 + z) z dL (z) = , (5.12) (0) 1/2 H0 ˜ )3 + (0) 0 [(1 − DE )(1 + z DE ] which can be evaluated numerically for given (0) DE . In order to satisfy the observational data (H0 /c)dL (z = 0.657) = 0.920 for 1997R, we require that (0) DE = 0.70. = 0.38 from the 1995ck data. Both data indicate the existence Similarly we get (0) DE of dark energy. Since observational data are prone to statistical and systematic errors, a few data are not enough to conclude that the present Universe is accelerating. Using 42 high-redshift SN Ia at redshifts between 0.18 and 0.83 together with 18 low-redshift SN Ia data from the Calan/Tololo Supernova Survey [106], Perlmutter et al. [2] showed that the cosmological constant is present at the 99% confidence level. They also found that the open Universe without the cosmological constant does not fit the data well. From Eq. (5.9) the apparent luminosity m gets larger for increasing luminosity distance dL . Figure 5.3 shows that the observational data in the highredshift regime favor the luminosity distance larger than the one predicted by the (0) CDM model ((0) m = 1 and = 0). From the likelihood analysis of the SN Ia data accumulated by the year 1998, Perlmutter et al. found that the density parameter of +0.09 non-relativistic matter is constrained to be (0) m = 0.28−0.08 (1σ statistical) in the flat Universe with the cosmological constant [2].
5.2 Supernova observations
91 (0)
26
(0.5, 0.5) (0, 0) (1, 0) (1, 0) (1.5, –0.5) (2, 0)
24
Flat
Supernova Cosmology Project
22
effective m B
(0)
20
18
Calan/Tololo (Hamuy et al., A.J. 1996)
16
14 0.02
0.05
0.1
0.2
0.5
1.0
redshift z Figure 5.3 The effective apparent luminosity mB versus the redshift z for 42 highredshift SN Ia from the SCP [2] and 18 low-redshift SN Ia from the Calan/Tololo Supernova Survey [106]. Note that mB involves the corrections of both sets for the SN Ia light-curve width-luminosity relation to the apparent luminosity m. The inner and outer error bars represent the uncertainty of measurements and the total uncertainty when the luminosity dispersion, 0.17 mag, of light-curve-widthcorrected SN Ia is added in quadrature, respectively. The horizontal error bars show the assigned peculiar velocity uncertainty of 300 km s−1 . The solid curves are the theoretical prediction for mB for a number of cosmological models without (0) the cosmological constant: ((0) m , ) = (0, 0) (top), (1, 0) (middle), and (2, 0) (bottom). The dashed curves correspond to a number of flat cosmological models: (0) ((0) m , ) = (0, 1) (top), (0.5, 0.5) (second from top), (1, 0) (third from top), and (1.5, −0.5) (bottom). From Ref. [2].
92
Observational evidence of dark energy 0.0 Union08 Riess Gold
Union08 Davis 07
Union08 Union08 w/o NB99
w
wDE
-0.5
-1.0
0.0
0.1
0.2
0.3
(0) m
0.4
-1.5 0.0
0.1
0.2
0.3
(0) m
0.4
0.0
0.1
0.2
0.3
(0)
0.4
0.5
m
Figure 5.4 68.3%, 95.4%, and 99.7% confidence level contours on ((0) m , wDE ) from the SN Ia observations only for constant wDE . In each column the filled contours correspond to the results coming from the Union data by Kowalski et al. [112]. The empty contours in the left and middle columns represent the constraints from the Gold sample [108, 109] and the ESSENCE data [110, 111], respectively. The right column shows the impact of the SCP nearby 1999 data. From Ref. [112].
After 1998 more SN Ia data have been collected by a number of high-redshift surveys – including SuperNova Legacy Survey [107] (SNLS), Hubble Space Telescope [108, 109] (HST), and “Equation of State: SupErNovae trace Cosmic Expansion” [110, 111] (ESSENCE) survey. The SNLS project, which is based on the Canada-France-Hawaii Telescope, consists of two components: (i) a large imaging survey to detect about 2000 supernovae and monitor their light curves, and (ii) a large spectroscopic survey to obtain supernovae identification and redshift. The HST survey is based on the image subtraction to search the SN Ia data in the high-redshift region z > 1 by including search depth, efficiency, timing, and falsepositive discrimination. These data have been classified as the “Gold” data sets [109]. The ESSENCE project is a ground-based survey designed to detect about 200 SN Ia in the redshift range z = 0.2–0.8 to measure the equation of state of dark energy to better than 10%. In Fig. 5.4 the observational contours on ((0) m , wDE ) are plotted from the SN Ia data by Kowalski et al. [112]. Note that the equation of state of dark energy is assumed to be constant. While the SN Ia data alone are not yet sufficient to place tight bounds on wDE , Fig. 5.4 clearly shows the presence of dark energy responsible for the late-time cosmic acceleration (wDE < −1/3). If the equation of state of dark energy varies in time, we need to parametrize wDE as a function of the redshift z. If the SN Ia data are accurate enough to measure the luminosity distance dL (z) in terms of z, it is possible to determine the evolution of
5.3 Cosmic Microwave Background
93
wDE by using Eqs. (2.86) and (2.88). However the error bars in the SN Ia data are still too large to determine wDE (z) without assuming some form of parametrization with respect to z. In particular, there is very little information on wDE at z > 0.5 [112]. The parametrization of wDE (z) will be discussed in Section 7.5. 5.3 Cosmic Microwave Background The observations of temperature anisotropies in the CMB provide another independent test for the existence of dark energy. The oldest sky we can see is the so-called last scattering surface at which electrons are trapped by hydrogen to form atoms (dubbed “decoupling” or “recombination”). The photons were tightly coupled to baryons and electrons before the decoupling epoch at z 1090, but they could freely move to us after that. In 1963 Penzias and Wilson [113] first detected the CMB photons thermalized to an almost uniform temperature across the sky. The temperature anisotropies of the CMB were first measured at large angular separations by the COBE satellite in 1992 [12]. The precise measurement of temperature anisotropies by high-precision experiments like BOOMERanG [114], MAXIMA [115], and especially WMAP [13] opened up a new opportunity to determine cosmological parameters to high precision. All of the matter components in the Universe (dark matter, neutrinos, . . . ) are coupled to gravity through the Einstein equations. The scalar part of the perturbations is the main source for the CMB temperature anisotropies. As we already mentioned the vector perturbations decay in the expanding Universe, whereas the tensor perturbations contribute to the CMB anisotropies as gravitational waves [90, 92]. However the amplitude of gravitational waves is suppressed relative to that of scalar perturbations if they originate from inflation (see Ref. [94]). Since the main feature of the CMB anisotropies is determined by scalar metric perturbations, we focus on scalar perturbations in the following discussion. Numerical simulations are required to derive the complete spectra of temperature and polarization anisotropies because the Einstein and Boltzmann equations for photons as well as other matter components are coupled to each other. In spite of this complexity, Hu and Sugiyama [116, 117] obtained many fitting formulae which are very helpful to understand the CMB physics and also derived analytic solutions under some approximations. In the following we shall discuss how the presence of dark energy affects the CMB anisotropies. The first effect is the change of the position of acoustic peaks coming from the modification of the angular diameter distance. The second effect is the so-called integrated Sachs–Wolfe (ISW) effect caused by the variation of the gravitational potential [118] (see Section 4.11.1). Since the latter is limited to very large scales, the first effect is typically more important.
94
Observational evidence of dark energy
We have already derived in Section 4.9 the first-order equation for the directionaveraged (i.e. monopole) temperature anisotropy 0 in the tight coupling regime (τop 1): 0 +
Rs Rs k2 H 0 + k 2 cs2 0 = − − H − , 1 + Rs 3 1 + Rs
(5.13)
where the sound speed squared is given by cs2 = 1/3(1 + Rs ) with the baryon-tophoton density ratio Rs = 3ρb /4ργ . This ratio evolves as Rs =
3ωb 1 , 4ωγ 1 + z
(5.14)
where we have defined 2 ωb ≡ (0) b h ,
2 ωγ ≡ (0) γ h .
(5.15)
If we take the value ωb = 0.02267 and ωγ = 2.469 × 10−5 we have Rs = 2.7 × 104 /(1 + z), which is much smaller than unity for redshifts larger than 105 (corresponding to the regime of tight coupling between photons and baryons). Taking the derivative of Rs with respect to η and using the definition of the redshift (2.28), we obtain the relation Rs = HRs . Then Eq. (5.13) can be written as 2 1 d Rs d k2 2 2 + − . (5.16) + k cs ( 0 + ) = dη2 1 + Rs dη 3 1 + Rs The second term on the l.h.s. of Eq. (5.16) is of the order of (Rs /η2 )( 0 + ) ≈ Rs H2 ( 0 + ), whereas the third term is of the order of k 2 cs2 ( 0 + ). Hence the second term can be neglected under the condition Rs cs2 (k/H)2 . In the tightcoupling regime (Rs 1) this condition is well satisfied for the modes inside the solution to Eq. (5.16) can be Hubble radius (k > H). Hence the homogeneous described by the sum of the solutions exp(±i kcs dη), i.e. ( 0 + )(hom) (k, η) = c1 f1 (k, η) + c2 f2 (k, η) ,
(5.17)
where f1 (η) = sin[krs (η)], f2 (η) = cos[krs (η)], and rs is the sound horizon defined by η dη˜ cs (η) ˜ . (5.18) rs (η) ≡ 0
It is clear from Eq. (5.17) that the homogeneous solution leads to oscillations in CMB anisotropies with a time-dependent frequency ωk (η) = krs (η).
5.3 Cosmic Microwave Background
95
The general solution to Eq. (5.16) can be obtained by the usual Green’s function method. In the tight-coupling regime (Rs 1) it is given by ( 0 + )(k, η) = c1 f1 (η) + c2 f2 (η) ˜ 2 (η) − f1 (η)f2 (η) ˜ k2 η f1 (η)f + . dη˜ [(η) ˜ − (η)] ˜ 3 0 f1 (η)f ˜ 2 (η) ˜ − f1 (η)f ˜ 2 (η) ˜ (5.19) We take the initial conditions 0 = 0 and = 0 at η = 0, which determine the coefficients c1 and c2 . Then Eq. (5.19) reduces to ( 0 + )(k, η) = [ 0 (0) + (0)] cos(krs ) η k dη˜ [(η) ˜ − (η)] ˜ sin [k(rs (η) − rs (η))] ˜ . +√ 3 0
(5.20)
Using Eqs. (4.176) and (5.20), the solution to the dipole moment is given by 1 1 (k, η) = √ [ 0 (0) + (0)] sin(krs ) 3 k η − dη˜ [(η) ˜ − (η)] ˜ cos [k(rs (η) − rs (η))] ˜ . 3 0
(5.21)
The first term on the r.h.s. of Eq. (5.20) shows that there is a peak for the homogeneous solution of 0 + at the position satisfying the relation krs = nπ ,
(5.22)
where n are integers. The peak position is subject to change by the presence of the last term in Eq. (5.20). The dipole solution (5.21) also contributes to the CMB spectrum. Recall that Eqs. (5.20) and (5.21) have been derived in the tight-coupling limit (τop 1). In order to obtain the CMB anisotropies observed today, we need to take into account the evolution of perturbations after the photons began to stream freely. (i.e. the epoch in which the tight coupling approximation is no longer valid). Moreover the contribution of higher-order moments ( with ≥ 2) and the ISW effect modifies the shape of the power spectrum. Hu and Sugiyama [116] derived the following semi-analytic expression for the present temperature field (see also
96
Observational evidence of dark energy
[74] for the derivation): (k, η0 ) [ 0 (k, η∗ ) + (k, η∗ )] j [k(η0 − η∗ )] ( + 1)j [k(η0 − η∗ )] + 3 1 (k, η∗ ) j−1 [k(η0 − η∗ )] − k(η0 − η∗ ) η0 ! dηe−τop (k, η) − (k, η) j [k(η0 − η)] , (5.23) + 0
where j (x) are spherical Bessel functions (see the Mathematical Appendix in Chapter 17) and η∗ is the time at which the visibility function −τop g(η) ≡ −τop e
(5.24)
takes a peak value. At the early stage much before the decoupling epoch (where the optical depth τop is much larger than 1), the visibility function g(η) is nearly ) zero. It takes a peak value roughly around z ∼ 1000. The scattering rate (−τop decreases rapidly after the decoupling epoch and hence g(η) approaches 0 again. The analytic solution (5.23) reproduces numerical solutions within 10% accuracy. The first term on the r.h.s. of Eq. (5.23) comes from the integral η0 dη g(η) [ 0 (k, η) + (k, η)] j [k(η0 − η)]. This carries the information of the 0 monopole solution (5.20) derived under the tight-coupling approximation. The effect of free streaming of photons appears in the visibility function g(η) whose main contribution to the integral comes from the value at η = η∗ [note that g(η) η defined in Eq. (5.24) satisfies the normalization condition 0 0 dη g(η) = 1]. The position of acoustic peaks estimated by Eq. (5.22) is shifted toward larger scales by the effect of free streaming of photons [74]. The second term in Eq. (5.23) coming from the dipole contribution 1 also leads to the modification of the CMB power spectrum. The third term in Eq. (5.23) is responsible for the so-called ISW effect, which is induced by the variation of the gravitational potentials and . While and remain nearly constants during the matter-dominated epoch, the dominance of dark energy at late times gives rise to their variations. We have already discussed this issue in Section 4.11.1. In the limit 1 the spherical Bessel function j (x) has a dependence j (x) (1/)(x/)−1/2 , which is suppressed for large . Hence the dominant contribution to the ISW effect corresponds to the low modes ( = O(1)). When we confront the predicted temperature anisotropies with CMB observations, we expand the perturbation in terms of spherical harmonics: (x, η) =
∞
ˆ , am (x, η)Ym (n)
(5.25)
=1 m=−
where the subscripts and m are conjugate to a real space unit vector nˆ representing the direction of incoming photons. The spherical harmonics Ym satisfies the
5.3 Cosmic Microwave Background
97
Multipole moment l 10
100
500
1000
6000
5000
4000 3000 2000 1000
0 90°
2°
0.5°
0.2°
Angular size Figure 5.5 The CMB power spectrum ( + 1)C /2π versus the multiple moment and the angular size θ . The relation between and θ is θ = π/. The curve shows the theoretical prediction of the power spectrum, whereas the black points present the WMAP 5-year data. From the webpage of WMAP: http://map.gsfc.nasa.gov/
normalization condition
ˆ ∗ m (n) ˆ = δ δmm . d Ym (n)Y
(5.26)
The coefficients am in Eq. (5.25) are assumed to be statistically independent. This means that the mean value of am ’s is zero (am = 0) with a non-zero variance defined by C ≡ |am |2 .
(5.27)
The variance C can be expressed in terms of the temperature field (k) in Fourier space (this equation will be derived in Section 14.1.1): 2 ∞ C = dk k 2 | (k)|2 . (5.28) π 0 In Fig. 5.5 we show the predicted CMB temperature anisotropies ( + 1)C /2π versus the multipole moment together with the WMAP 5-year observational data [15]. The theoretical power spectrum agrees with the observational data with flying colors. The measured angle θ has a relation θ = π/ [rad] with the multipole .
98
Observational evidence of dark energy
Hence the larger scales correspond to lower values of . The large-scale power spectrum ( 10) is dominated by the monopole mode 0 , which inherits the information of nearly scale-invariant density perturbations generated during inflation. The reason why the quantity ( + 1)C /2π is plotted instead of C itself is that the former is constant for scale-invariant perturbations on large scales, 2 2 , where δH is the amplitude of curvature perturbations i.e. ( + 1)C = (π/2)δH generated during inflation [74]. Equation (5.22) shows that the comoving wavelength corresponding to acoustic peaks can be approximately estimated as λc = 2π/k = (2/n)rs . We then define the following characteristic angle for the location of peaks: θA ≡
rs (zdec ) dA(c) (zdec )
,
(5.29)
where zdec is the redshift at the decoupling epoch and dA(c) is the comoving angular diameter distance defined by dA(c) (z) ≡
dA (z) = (1 + z)dA (z) . a
(5.30)
The physical (proper) diameter distance dA (z) is given by Eq. (2.73). The multipole corresponding to the angle (5.29) is A =
π d (c) (zdec ) . =π A θA rs (zdec )
(5.31)
From Eqs. (2.73) and (5.30) the comoving angular diameter distance dA(c) (zdec ) is expressed as dA(c) (zdec ) =
1 c R, H0 (0) m
where R is the so-called CMB shift parameter defined by . (0) zdec . m dz . R = / (0) sinh (0) K E(z) K 0 From Eqs. (4.180) and (5.18), the sound horizon rs (zdec ) is ∞ dz c . rs (zdec ) = √ √ 3a0 H0 zdec 1 + Rs E(z) For the redshift zdec there is a fitting formula by Hu and Sugiyama [117] g2 zdec = 1048 1 + 0.00124ωb−0.738 1 + g1 ωm ,
(5.32)
(5.33)
(5.34)
(5.35)
5.3 Cosmic Microwave Background
99
2 where ωm ≡ (0) m h and
g1 = 0.0783ωb−0.238 / 1 + 39.5ωb0.763 ,
g2 = 0.560/ 1 + 21.1ωb1.81 . (5.36) The bounds coming from the WMAP 5-year data correspond to ωb = 0.02265 ± 0.00059 and ωm = 0.1369 ± 0.0037. Taking the values ωb = 0.02265 and ωm = 0.1369, we obtain zdec = 1090.98 from Eq. (5.35). Since the contribution of dark energy to E(z) in Eq. (5.34) is negligible for √ 2 z > zdec , one can estimate this quantity to be E = ( a + aeq /a ) (0) m , where −1 aeq = (1 + zeq ) is the scale factor at the radiation–matter equality [see Eq. (2.78)]. Then the integral (5.34) can be written as adec 1 1 1 c rs (zdec ) = √ da , (5.37) √ √ 1 + Rs (a) a + aeq 3H0 (0) 0 m where Rs (a) = (3ωb /4ωγ ) a . Equation (5.37) is integrated to give " (eq) (dec) (dec) ωγ 4 ch Rs + Rs + 1 + Rs ln rs (zdec ) = , 3 H0 ωm ωb (eq) 1 + Rs
(5.38)
(5.39)
(eq)
where Rs(dec) ≡ Rs (adec ) and Rs ≡ R(aeq ). From Eqs. (5.32) and (5.39) the multipole A in Eq. (5.31) is −1 (eq) (dec) (dec) " 3π ωb Rs + Rs + 1 + Rs A = R ln , 4 ωγ (eq) 1 + Rs
(5.40)
which shows that A is proportional to R. The CMB shift parameter defined in Eq. (5.33) is affected by the cosmic expansion history from the decoupling to the present. The presence of dark energy leads to a shift of R compared to the CDM model, thereby changing the value of A . Hence the CMB shift parameter can be used to place constraints on dark energy. The bound on R according to the WMAP 5-year data is given by R = 1.710 ± 0.019
(68% confidence level) .
(5.41)
For example, let us compute the multipole A for R = 1.710, ωb = 0.02265, ωm = 0.1369, and ωγ = 2.469 × 10−5 . Using Eqs. (2.77), (5.35), and (5.38) we obtain the value A = 299. This is different from the location of the first acoustic peak
100
Observational evidence of dark energy 2.0
CMB shift parameter R
wDE = −1 1.8
WMAP bound 1.6
1.4
wDE = −0.5 1.2
1.0 0
0.2
0.4
0.6
0.8
1
Ω DE(0) Figure 5.6 The CMB shift parameter R versus (0) DE for wDE = −1 and wDE = −0.5 in the flat Universe. We also show the bound coming from the WMAP 5-year observational data. For the cosmological constant (wDE = −1) the observational (0) constraint on (0) DE coming from the CMB shift parameter is 0.72 < DE < 0.77.
which is located around 1 = 220 (see Fig. 5.5). As we already explained, this shift comes from several effects such as the free streaming of photons and the dipole contribution. We write the general relation for all peaks and troughs of observed CMB anisotropies as m = A (m − φm ) ,
(5.42)
where m represents peak numbers (m = 1 for the first peak, m = 1.5 for the first trough, . . . ) and φm is the shift of multipoles. It is known that φm depends weakly on ωb and ωm for a given cosmic curvature (0) K . According to the fits by Doran and Lilley [119] the shift of the first peak is about φ1 = 0.265, which gives 1 = 220 for A = 299. In the flat Universe the CMB shift parameter (5.33) reduces to zdec dz , (5.43) R = (0) m E(z) 0 (0) 3 (0) 4 3(1+wDE ) 1/2 where E(z) = [(0) ] for the conm (1 + z) + r (1 + z) + DE (1 + z) stant dark energy equation of state. In Fig. 5.6 we plot R versus (0) DE for the cases wDE = −1 and wDE = −0.5 together with the WMAP bound (5.41). For larger
5.3 Cosmic Microwave Background
101
wDE
−0.5
−1.0 WMAP WMAP+HST
−1.5
WMAP+BAO WMAP+SN WMAP+BAO+SN
−2.0 0.4
0.5
0.6 (0) Ω DE
0.7
0.8
Figure 5.7 Observational contours (68% and 95% confidence level) on the constant dark energy equation of state wDE and the present dark energy density (0) DE for the flat Universe. The WMAP 5-year data alone do not provide strong constraints on wDE and (0) DE . This is slightly improved by adding the measurement of H0 from the Hubble key result. The observational contours coming from WMAP+BAO, WMAP+SN Ia, and WMAP+BAO+SN Ia are also shown. The combined analysis of WMAP+SN Ia data provide a tight constraint −1.098 < wDE < −0.841 at the 95% confidence level. From Ref. [15].
(0) DE the CMB shift parameter (5.43) gets smaller. When wDE = −1 the density parameter is constrained to be 0.72 < (0) DE < 0.77 from the bound (5.41). From Fig. 5.6 we find that the observationally allowed values of (0) DE become smaller for increasing wDE . Since the CMB shift parameter depends only weakly on wDE , the equation of state wDE is not strongly constrained by the CMB data alone. In Fig. 5.7 the combined observational bounds on (0) DE and (constant) wDE are plotted in the flat Universe. This is the joint analysis derived by using the observational data of WMAP 5-year, SN Ia, and BAO. As expected, the WMAP 5-year data alone do not provide strong constraints on wDE , although the evidence for the accelerated expansion (wDE < −1/3) can be seen in Fig. 5.7. Notice that the WMAP bound in Fig. 5.7 is consistent with the bound coming from the CMB shift parameter discussed above. For decreasing wDE the observationally allowed range of (0) DE shifts to larger values. The prior on the Hubble constant H0 = 72 ± 8 km sec−1 Mpc−1 from the Hubble Key Project [72] improves the constraint a bit: −1.47 < wDE < −0.58 at the 95% confidence level. Adding the
102
Observational evidence of dark energy 3 0.04
1
0.03 0.02
0.3
0.01
ξ(s)
0 0.1
−0.01 50
100
150
0.04 0.02 0.00 −0.02
50 100 150 Comoving separation (h−1 Mpc)
Figure 5.8 The large-scale redshift-space correlation function of the SDSS sample. The inset shows an expanded view with a linear vertical axis. From top to bottom (0) 2 2 each curve corresponds to (0) m h = 0.12, 0.13, 0.14 with b h = 0.024 and a (0) 2 pure CDM (no baryons) model with m h = 0.15. The observational data clearly show the existence of an acoustic peak around the comoving separation scale 100 h−1 Mpc, in agreement with the predictions except for the pure CDM model. From Ref. [68].
“Union” SN Ia data by Kowalski et al. [112] significantly improves the constraint: −1.098 < wDE < −0.841 at the 95% confidence level. This shows the importance of combined analysis using independent observational data. In the next subsection we shall proceed to the constraint coming from baryon acoustic oscillations.
5.4 Baryon acoustic oscillations Since baryons are tightly coupled to photons before the recombination epoch, the oscillations of sound waves should be imprinted in the baryon perturbations as well as the CMB temperature anisotropies. Eisenstein et al. [68] found a peak of baryon acoustic oscillations in the large-scale correlation function at 100 h−1 Mpc separation measured from a spectroscopic sample of 46,748 luminous red galaxies from the Sloan Digital Sky Survey (SDSS), see Fig. 5.8. This detection of baryon oscillations provided another independent test for constraining the property of dark energy, which contributes to breaking residual degeneracies in the CMB data [121].
5.4 Baryon acoustic oscillations
103
At the tight coupling epoch in which baryons are strongly coupled to photons the perturbations in baryons, δb ≡ δρb /ρb , satisfy the following equation of motion δb +
Rs 3Rs Hδb + k 2 cs2 δb = −k 2 − H − 3 . 1 + Rs 1 + Rs
(5.44)
If we substitute the relation δb = 3 0 into Eq. (5.44), we obtain the same equation as Eq. (5.13) for the temperature field 0 . Since 0 = (1/4)δγ , this corresponds to 3 δb = δ γ , 4
(5.45)
which is the adiabatic condition we have already mentioned after Eq. (4.198). This is also associated with the fact that the ratio nb /s of the number density of baryons (nb ) to the entropy density (s = (ρ + P )/T ) does not vary in time. During the radiation era the entropy density is dominated by the contribution of relativistic 3/4 particles, which gives s ∝ T 3 ∝ ργ . Using the relation ρb ∝ nb , we find δnb δ(nb /s) δs 3 = = δb − δ γ , − nb /s nb s 4
(5.46)
which vanishes under the condition (5.45). Since the baryon–entropy ratio is independent of the temperature T , it takes a constant value at different spatial positions when the perturbations are generated. Hence the condition δb = (3/4)δγ sets the (adiabatic) initial condition for baryon perturbations. This relation is preserved in the tight-coupling era so that baryons and photons evolve as a single fluid. The sound horizon at which baryons were released from the Compton drag of photons plays a crucial role to determine the location of baryon acoustic oscillations. This epoch, called the drag epoch, occurs at the redshift zdrag . The sound horizon at z = zdrag is ηdrag rs (zdrag ) = dη cs (η) , (5.47) 0
where cs is given in Eq. (4.180). We caution that the drag epoch does not coincide with the recombination epoch at which photons were released from electrons. For the redshift zdrag there is a fitting formula by Eisenstein and Hu [120]: zdrag =
0.251 1291ωm 1 + b1 ωbb2 , 0.828 1 + 0.659ωm
(5.48)
where −0.419 0.674 b1 = 0.313ωm 1 + 0.607ωm ,
0.223 . b2 = 0.238ωm
(5.49)
104
Observational evidence of dark energy
The WMAP 5-year data constrain the values of zdrag and rs (zdrag ) to be zdrag = 1020.5 ± 1.6 and rs (zdrag ) = 153.3 ± 2.0 Mpc [15]. We observe the angular and redshift distributions of galaxies as a power spectrum P (k⊥ , k ) in the redshift space, where k⊥ and k are the wavenumbers perpendicular and parallel to the direction of light respectively. As we will show in more detail in Section 14.2, it is possible to measure the following two ratios [122] rs (zdrag ) , (1 + z)dA (z) rs (zdrag )H (z) , δzs (z) = c
θs (z) =
(5.50) (5.51)
where the denominator in Eq. (5.50) corresponds to the comoving diameter distance dA(c) (z) = (1 + z)dA (z) defined in Eq. (5.30). The angle θs (z) is completely analogous to the CMB acoustic peak angle introduced in Eq. (5.29) and corresponds to observations orthogonal to the line of sight. The quantity δzs is instead measured by identifying in the fluctuation spectrum the oscillations along the line of sight. So far the BAO data have not been accumulated sufficiently to measure the two distances θs (z) and δzs (z) separately [123]. However it is possible to obtain a combined distance scale ratio from the spherically averaged spectrum: !1/3
θs (z)2 δzs (z)
≡
[(1 +
rs (zdrag ) 2 z) dA2 (z)c/H (z)]1/3
or the related effective distance [68] cz 1/3 2 2 , DV (z) ≡ (1 + z) dA (z) H (z)
,
(5.52)
(5.53)
obtained from the combination of two spatial dimensions orthogonal to the direction of sight and one dimension along the direction of sight. The current constraint from SDSS data is DV (z = 0.35) = 1370 ± 64 Mpc at a typical redshift z = 0.35 [68]. Another observational constraint comes from the 2-degree Field (2dF) Galaxy Redshift Survey. Defining the relative BAO distance rBAO (z) ≡ rs (zdrag )/DV (z) ,
(5.54)
we have at the two redshifts (z = 0.2 and z = 0.35) [124] rBAO (z = 0.2) = 0.1980 ± 0.0058 ,
rBAO (z = 0.35) = 0.1094 ± 0.0033 . (5.55)
5.4 Baryon acoustic oscillations
105
0.30
(i)
(ii)
0.20
0.15
.
r BAO (z)
(iii)
.
0.25
0.10
0.05
0.0 0
0.1
0.2
0.3
0.5
0.4
z Figure 5.9 The BAO distance ratio (5.54) versus the redshift z in the flat Universe (K = 0) with the cosmological constant (wDE = −1) for three different cases: (0) (0) (i) (0) DE = 0.75, (ii) DE = 0, and (iii) DE = 0.95. We choose the values ωγ = −5 2.469 × 10 and ωb = 0.02265. We also show the most likely observational values of rBAO (z) at z = 0.2 and z = 0.35 as black points. The observational data favor the dark energy density around (0) DE = 0.75.
The sound horizon rs (zdrag ) can be obtained by replacing Rs(dec) in Eq. (5.39) for (drag) Rs ≡ Rs (zdrag ), whereas the angular diameter distance dA is given in Eq. (2.73). This gives the explicit form for rBAO (z):
rBAO (z) =
"
−1/3
(0) sinh K
1 (0) K (drag) (drag) (eq) Rs + Rs + 1 + Rs . × ln (eq) 1 + Rs 4 3
ωγ
(0) m ωb
z E(z)
z 0
−2/3
d˜z E(˜z)
(5.56)
In Fig. 5.9 we plot rBAO (z) versus z for the CDM model in the flat Universe (wDE = −1 and K = 0). This shows that the CDM model with (0) DE = 0.75 is favored over the CDM model from the BAO data at z = 0.2 and z = 0.35. If the equation of state of dark energy wDE is not −1 but a constant, the models with wDE < −1, dubbed “phantoms,” are favored by the present BAO data. This
106
Observational evidence of dark energy 105 ACDM CDM
104
103
102 0.001
0.01
0.1
1
Figure 5.10 The predicted matter power spectra Pδm in the flat Universe for (0) the CDM model with (0) m = 0.28 and the CDM model with m = 1.0. We 2 = use the expression (4.212) of the power spectrum with model parameters δH −10 (0) 3.2 × 10 , ns = 0.96, and h = 0.7. For smaller m the position of the peak is shifted toward smaller k.
can be seen in the combined analysis of WMAP 5-year and BAO data in Fig. 5.7, which shows that even the equation of state such as wDE = −1.5 can be allowed. However the addition of the SN Ia data provides a tight constraint: −1.097 < wDE < −0.858 at the 95% confidence level. This constraint is not much different from the joint analysis of the WMAP+SN Ia data without the BAO data (−1.098 < wDE < −0.841 at the 95% confidence level). The non-linear evolution of baryon acoustic oscillations was extensively studied in Refs. [125, 126] by comparing N -body simulations with analytically estimated spectra for the CDM model.
5.5 Large-scale structure The observations of large-scale structure such as the galaxy clustering provides another test for the existence of dark energy. The wavenumber at the peak position of Pδm corresponds to keq given in Eq. (4.195). This shows that keq decreases for smaller values of (0) m . In Fig. 5.10 we plot the predicted matter power spectrum for (0) the flat CDM model ((0) m = 0.28) and for the CDM model (m = 1.0). In the presence of dark energy the peak position shifts toward larger scales (i.e. for smaller
5.5 Large-scale structure
107
LRG Main
Power spectrum P(k) [(h−1Mpc)3]
105
104
0.01
k [h Mpc−1]
0.1
Figure 5.11 Measured power spectra with error bars for the full luminous red galaxies (LRG) and main galaxy samples of the 2dF survey. The solid curves show the theoretically predicted spectra for the CDM model obtained in the linear perturbation theory with galaxy bias b = 1.9 (top) and b = 1.1 (bottom) relative to the z = 0 matter power spectrum. The dashed curves include the non-linear correction to the matter spectrum by Cole et al. [88]. The non-linear effect becomes important for the scales k 0.09h Mpc−1 . From Ref. [127]. (0) −1 k). For (0) and m = 1 and m = 0.28 with h = 0.7 we have keq = 0.051 h Mpc −1 keq = 0.014 h Mpc , respectively. Hence the scale of the peak position can be used as a probe of dark energy. As we have seen in Section 4.8 the matter spectrum Pδm is related to the observed galaxy power spectrum Pδg via the relation
Pδg = b2 Pδm ,
(5.57)
where b is the bias parameter. In Fig. 5.11 the galaxy power spectra of luminous red galaxies (LRG) and main galaxy samples of the SDSS are plotted [127]. The position of the peak, around the scale 0.01 h Mpc−1 < k < 0.02 hMpc−1 , shows that the CDM model is favored over the CDM model. Although the galaxy power spectra alone do not provide tight bounds on the density parameter (0) DE as well as wDE , the important point is that the observations of LSS are consistent with the existence of dark energy. Notice that the linear spectrum plotted in Fig. 5.10
108
Observational evidence of dark energy
is modified for the scales k 0.09 h Mpc−1 because non-linear effects become important on smaller scales [88]. 5.6 Problems 5.1 Integrate Eq. (5.12) numerically for four different values (0) DE = 0, 0.3, 0.7, 1 and plot −1 the luminosity distance dL (in units of cH0 ) in the flat Universe as a function of z. (0) 5.2 Plot the BAO distance ratio (5.56) versus the redshift z for (0) DE = 0, DE = 0.75, and (0) DE = 0.95.
6 Cosmological constant
The simplest candidate for dark energy is the cosmological constant , which is so called because its energy density is constant in time and space. In fact the CDM model has been systematically proved consistent with a large number of observations. The Lagrangian density for the CDM model is simply given by the linear term in R plus , see Eq. (6.2). Despite its simplicity it is generally difficult to explain why the energy scale of the cosmological constant required for the cosmic acceleration today is very small relative to that predicted by particle physics. As we already mentioned, the vacuum energy density evaluated by summing the zero-point energy of a scalar field is about 10121 times larger than the observed dark energy density (for a momentum cut-off around the Planck scale). The problem of a large value of was present long before the observational discovery of the late-time cosmic acceleration. In fact, even if we had no observational evidence of dark energy we would still need to understand why the cosmological constant vanishes. Models of dark energy alternative to CDM are based on the assumption that is zero or negligible. So the problem of the cosmological constant is to find some mechanism that either makes it vanish or renders it a very small value compatible with the present cosmological density. In the former case the origin of dark energy needs to be explored further, but in the latter case the problems of the cosmological constant and dark energy are solved at the same time. In this chapter we first present the action for the CDM model and show that the Einstein equations follow from this by the action principle. We then proceed to the history of the cosmological constant and its fine-tuning problem. We review a number of attempts to solve the cosmological constant problem in the framework of supergravity and superstring theories. We also discuss several topics related to the cosmological constant – such as the anthropic selection, and the decoupling of from gravity.
109
110
Cosmological constant
6.1 Einstein equations with the cosmological constant The energy-momentum tensor Tµν on the r.h.s. of the Einstein equations (2.8) obeys the conservation law Tµν ;ν = 0. Since the metric gµν satisfies the relation gµν ;ν = 0, it is possible to add the term gµν to the Einstein equations: 1 Rµν − gµν R + gµν = 8π GTµν , 2
(6.1)
where is the cosmological constant. It is interesting to note that these are the most general equations of second-order in the metric in four dimensions. In scalartensor metric theories an additional term coupled to a Gauss–Bonnet term is also allowed. The Einstein equations (6.1) can be derived by the action principle. It is based on the linear action in terms of the Ricci scalar R = g µν Rµν and the matter action Sm : √ 1 S= d4 x −g (R − 2) + Sm . (6.2) 16π G Let us derive the Einstein equations (6.1) from the action (6.2). It is important to understand this procedure because similar steps are taken when we consider modified gravity models in Chapter 9. The variation of the action (6.2) with respect to g µν gives √ 1 δS = d4 x δ( −g)(g µν Rµν − 2) 16π G ! √ √ + −g δg µν Rµν + −g g µν δRµν + δSm . (6.3) α α α ν );α − (δµα );ν we have g µν δRµν = (g µν δµν − g µα δµν );α Since δRµν = (δµν and hence √ 4 √ µν α ν d x −g g δRµν = d4 x −g(g µν δµν − g µα δµν );α = 0 , (6.4)
where we have employed Gauss’s theorem in the last equality. This shows that the last term in the square bracket of Eq. (6.3) vanishes. Now we also use the relation √ √ δ( −g) = −(1/2) −ggµν δg µν . This can easily be derived by differentiating with respect to g µν the determinant g written as gµν M(µν) where M(µν) is the determinant of the cofactor matrix, which does not depend on the element gµν itself, and then replacing M(µν) = gg µν . Then Eq. (6.3) reads 1 1 4 √ (6.5) δS = d x −g Rµν − Rgµν + gµν δg µν + δSm . 16π G 2
6.2 History of the cosmological constant
111
The energy-momentum tensor Tµν is defined from the variation of δSm in terms of g µν : √ 1 (6.6) δSm = − d4 x −g Tµν δg µν . 2 Then Eq. (6.5) reduces to 1 1 4 √ d x −g Rµν − Rgµν + gµν − 8π GTµν δg µν . δS = 16π G 2
(6.7)
The Einstein equations (6.1) follow from the action principle, δS = 0. 6.2 History of the cosmological constant After Einstein constructed General Relativity in 1915–1916 [128], he tried to apply his theory to the Universe in 1917 [23]. In the absence of the cosmological constant it is obvious from Eq. (2.19) that the scale factor a can dynamically change in time (except in the case of a fluid at rest with a specific equation of state: w = P /ρ = −1/3). In the 1910s, however, Einstein believed that the Universe was static and introduced the cosmological constant to realize such a Universe. For the FLRW metric (2.1) the Einstein equations (6.1) read K 8π G ρ− 2 + , (6.8) 3 a 3 a¨ 4π G =− (ρ + 3p) + . (6.9) a 3 3 From Eq. (6.9) it is clear that works as a repulsive force against gravity at the background level. In the Universe dominated by a pressureless matter (P = 0), we find that the static Universe (a˙ = a¨ = 0) corresponds to H2 =
K , = . (6.10) 4π G a2 This equation, the first relativistic cosmology ever, shows that the density ρ in the Universe is determined by . Einstein believed that this solution (a “crazy idea” according to his own words in a letter to de Sitter) was a way to embody Mach’s idea of linking mass (ρ) to inertia, here represented by space-time geometry gµν . He thought that he could eventually show that matter was necessary to define a non-Minkowskian metric. However, the above static solution is unstable against perturbations of the density ρ as was later demonstrated by Lemaˆıtre. In fact, if /3 > (4π Gρ)/3, Eq. (6.9) shows that the Universe departs from the static point given in Eq. (6.10) with the growth of a. If /3 < (4π Gρ)/3 the Universe is also away from the static point with the decrease of a. Einstein did not realize ρ=
112
Cosmological constant
this instability since he did not write down the differential equation for a(t). Shortly after, in the same year 1917, de Sitter [129] found his accelerated solution √ H = /3 in empty space, paving the way to the dismissal of Mach’s principle in cosmology. At the same time, from 1910 to the mid 1920s, Slipher was observing the spectra of galaxies (spiral nebulae) and found most of them to be red-shifted. In 1922 Friedmann found the evolving solution that represents the expanding Universe [130]. In 1927 Lemaˆıtre [131] studied the relation between the observed results of the redshift and the homogeneous Universe dominated by a pressureless dust. In Lemaˆıtre’s model there are three distinct periods for the evolution of the Universe: r (i) A period of cosmic expansion (a ∝ t n with 0 < n < 1) from a point source during which the basic elements were formed. This corresponds to the expanding Universe dominated by matter (either the pressureless matter or the radiation). r (ii) A period of a very slow expansion (a ∝ constant) during which nebulae were formed. This resembles the static Universe proposed by Einstein. r (iii) A period of a fast expansion (a ∝ t n with n > 1) during which the recession of the √ nebulae is accelerating. This period can be realized by the de Sitter solution (H = /3) in the presence of the cosmological constant.
Lemaˆıtre’s model is the first “hot Big Bang” model, in which the matter density ρ goes to infinity as a → 0. Apart from the existence of the period (ii) Lemaˆıtre’s model describes well the evolution of the Universe even in the modern context. The loitering period (ii) should be replaced by a short transient period from the matter era to the accelerated epoch during which the system crosses the point a¨ = 0, while the nebulae formed during the matter-dominated epoch [the period (i)]. The period (iii) is exactly the phase of the late-time cosmic acceleration realized by the presence of . We can say that Lemaˆıtre, influenced by de Sitter’s accelerated solution, produced the first consistent dark energy model. In 1929 Hubble formulated Hubble’s law (2.32) by combining his measurements of galaxy distances with Slipher’s measurements of the redshifts associated with the galaxies [71]. This was the first direct quantitative evidence for the expansion of the Universe. The existence of the cosmological constant was clearly not required to give rise to a (decelerated) cosmic expansion. In the book “The Meaning of Relativity” written by Einstein in 1945 [132], he stated that “if Hubble’s expansion had been discovered at the time of the creation of the general theory of relativity, the cosmological member (the cosmological constant) would never have been introduced.” In 1970 Gamov [133] recalls that “when I was discussing cosmological problems with Einstein, he remarked that the introduction of the cosmological term was the biggest blunder he ever made in his life.” In spite of Einstein’s regret, the cosmological constant returned at the end of the century to account for the late-time cosmic acceleration.
6.3 The fine tuning problem
113
6.3 The fine tuning problem In order to realize the cosmic acceleration today, we require that the cosmological constant is of the order of the square of the present Hubble parameter H0 [see Eq. (6.8)]: ≈ H02 = (2.1332h × 10−42 GeV)2 .
(6.11)
If we interpret this as an energy density, it is equivalent to ρ ≈
m2pl 8π
≈ 10−47 GeV4 ≈ 10−123 m4pl ,
(6.12)
where we have used h ≈ 0.7 and mpl ≈ 1019 GeV. Suppose that the energy density (6.12) comes from the vacuum energy ρ of an empty space. The zero-point energy of√ some field of mass m with momentum k and frequency ω is given by E = ω/2 = k 2 + m2 /2 (in the units of = c = 1). Summing over the zero-point energies of this field up to a cut-off scale kmax ( m), we obtain the vacuum energy density kmax 3 d k 1 2 ρvac = k + m2 . (6.13) 32 (2π ) 0 Since the integral is dominated by the mode with large k ( m), we find that kmax k4 4π k 2 dk 1 2 2 ≈ max . k + m (6.14) ρvac = (2π )3 2 16π 2 0 General Relativity is believed to be valid up to the Planck scale mpl . Taking the cut-off scale kmax to be mpl , the vacuum energy density can be estimated as ρvac 1074 GeV4 .
(6.15)
This is about 10121 times larger than the observed value (6.12). Note that this situation is not improved much by taking other energy scales appearing in particle physics. For the QCD scale kmax ≈ 0.1 GeV we have ρvac ≈ 10−3 GeV4 , which is still much larger than ρ . The above problem was present even before the observational discovery of dark energy in 1998. At that time most people believed that the cosmological constant was exactly zero and tried to explain why it was so. The vanishing of a constant usually implies the existence of some symmetry. In supersymmetric theories, for example, the bosonic degree of freedom has its Fermi counterpart that contributes to the zero-point energy with an opposite sign. If supersymmetry is unbroken, there exists an equal number of bosonic and fermionic degrees of freedom such that the total vacuum energy vanishes [134] [see Eq. (6.22) in Section 6.5]. However it is known that supersymmetry is broken at sufficient high energies (around the scale MSUSY = 103 GeV if it is relevant to the hierarchy problem of gravitational
dN
Cosmological constant
dΩΛ
114
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 − 10
−5
0
5
10
log10 (a) Figure 6.1 Plot of d /dN versus log10 a, assuming flat space with (0) = 0.7. The spike is very close to the present epoch: this is the coincidence problem.
interaction and weak interaction). Hence the vacuum energy is generally non-zero in the world of broken supersymmetry. Nevertheless, it is not impossible to obtain a vanishing or a tiny amount of even if supersymmetry is broken. We shall address this problem in Section 6.5.
6.4 The coincidence problem The second problem of the cosmological constant as dark energy is that its value is not only at odds with all possible fundamental energy scales and requires therefore fine tuning, but also that this particular value is almost identical to a totally unrelated number, the present matter energy density. In other words, (0) is doubly unlikely: because it is too small in absolute terms and because its value coincides (to a factor of two or three) with (0) m , for no obvious reason. The matter density ρm = ρm(0) (1 + z)3 coincides with the cosmological density ρ(0) at 1/3 (0) zcoinc = − 1, (6.16) 1 − (0) which, for (0) = 0.7, amounts to zcoinc ≈ 0.3. This problem is called the coincidence problem. To illustrate the issue, we plot the evolution of the derivative d /dN (where N = ln a) in Fig. 6.1. We find that the only epoch in which this quantity is not close to zero is the present.1 If ρ(0) /ρm(0) was just 10 or 100 times smaller, we would not see any accelerated expansion. If it were a few orders of magnitude larger than 1
We note, however, that the choice of a natural time scale to define the coincidence problem is matter of debate, see e.g. Ref. [135].
6.4 The coincidence problem
115
unity, the spike would occur at a large redshift and probably we would not call it a coincidence at all. The coincidence problem is not specific to the cosmological constant. Almost all acceptable dark energy models we will see in the next chapters behave similarly to the cosmological constant and their zcoinc also turn out to be very close to zero. Therefore we discuss this problem in terms of a general dark energy density ρDE . Barring the case that this coincidence is after all just a coincidence, or that all the observational evidence in favor of acceleration is systematically wrong, cosmologists have proposed several ways out of this problem. The first class of explanations is based on models in which ρDE responds to the trend of ρm and catches up with it irrespective of the initial conditions of ρDE . In this case DE is non-zero for a considerable duration and this alleviates the coincidence problem. However, the acceleration starts very recently and therefore a coincidence arises again. The problem is in fact that this behavior is based on attractor-like solutions such as the so-called tracker models, see Section 7.2.3. The second class of explanation argues that there is no coincidence and in fact m and DE have always, or most of the time, been similar. In principle this is not difficult to realize: it is sufficient to postulate two components, one that clusters, the other that does not because of a large sound speed, and to regulate their equations of state so that they are always similar. The main problem here is that either (i) the common equation of state always satisfies the condition for cosmic acceleration and hence it is difficult or impossible to be consistent with many observations such as the growth of large-scale structure, or (ii) the equation of state changes right when it is needed, i.e. today, and therefore another coincidence arises – this time between the epoch of acceleration and the present. Models that belong to this class are for instance the scaling attractors, which will be discussed in Section 8.5.3. A related possibility is to build a model with several epochs of acceleration; it is then just a matter of reasonable chance to be witnessing one. Here again the difficulty is to realize a sufficient period of structure formation. The third class is the anthropic one. According to it we live in a Universe with ρDE ≈ ρm because this is the highest dark energy density allowed by the requisite of sufficient structure formation and, in general, higher energy vacua are more likely than lower ones. So our Universe is the most likely among the “life-sustaining” universes. We discuss anthropic arguments in Section 6.6. The fourth class is the “backreaction” argument. The coincidence between ρDE and ρm may appear as a by-product of another more fundamental one, the coincidence between acceleration and structure formation. This can be explained if one causes the other: in particular, if the growth of structures causes acceleration through cumulative non-linear effects. We discuss these models in Section 10.2. A related way out is that actually there is no real acceleration and no dark energy. The acceleration is only an apparent consequence of adopting the wrong background
116
Cosmological constant
cosmological model, the FLRW spacetime. If instead we interpret observations with a strongly inhomogeneous model like the Lemaˆıtre–Tolman–Bondi void, the acceleration of the recession rate between nearby and distant sources becomes a distance-dependent, but practically always decelerated, Hubble rate. We discuss this interpretation in Section 10.1. A quick summary of this section is that the coincidence problem is far from solved. It is difficult to imagine a convincing explanation of the nature of dark energy which does not at the same time provide a solution to the coincidence problem. Until then, we can use the coincidence problem, just as the fine-tuning problem, as a guide to select interesting directions of research. 6.5 Supersymmetric models From here to the end of this section we discuss the status of the cosmological constant in supersymmetric models. This material is more advanced than most of this book and requires knowledge of quantum field theory and supersymmetry. However it is not a necessary prerequisite for the subsequent chapters. Supersymmetric theoretical models consist of a set of quantum fields having a symmetry between bosons and fermions (see Ref. [136] for introductory review). In particular a supersymmetric model covariant under general coordinate transformations (or a model having local supersymmetry) is called a supergravity model. Supergravity aims to unify the gravitational force with other interactions. When we quantize gravity, the problem of renormalizability is crucial. In renormalizable theories the divergence of integrals which appears in the perturbative expansion for physical processes can be avoided by redefining a finite number of theoretical parameters. Unfortunately supergravity is a non-renormalizable theory and hence it should not be regarded as “a theory of everything.” Superstring theory is the only known theory containing gravity that is renormalizable. Supergravity can be regarded as an effective low-energy theory derived from more fundamental string theory. In the following we shall briefly review supersymmetric theories and then proceed to possible solutions to the cosmological constant problem. The readers who are interested in the details of supergravity and superstring theories may consult the books of Bailin and Love [137] and Green, Schwarz, and Witten [138]. The transformations of supersymmetry are based on quantum operators Qs which change bosonic states into fermionic ones and vice versa, Qs |boson = |fermion ,
Qs |fermion = |boson .
(6.17)
From this definition the Qs ’s change the spin of states. The Qs ’s are identified as spinor operators transforming like tensor operators of spin 1/2. It is known that the Qs ’s are invariant under coordinate translations. More technically, the Qs ’s commute with the four generators P µ of the translation group. We use the
6.5 Supersymmetric models
117
Hermitian generator for P µ = (E, P), so that E and P correspond to the energy operator and the momentum operator, respectively. Then the commutation relation is given by [Qs , P µ ] ≡ Qs P µ − P µ Qs = 0 .
(6.18) †
The anti-commutator of a generator Qs with its Hermitian adjoint Qs is a linear combination of energy and momentum operators: {Qs , Q†s } ≡ Qs Q†s + Q†s Qs = αE + β P .
(6.19)
If we sum up all supersymmetry generators, then the β P terms cancel whilst the αE terms add up. This then gives the following dependence {Qs , Q†s } = cE . (6.20) all Qs
Here the proportionality factor c is positive for a physically meaningful theory with energies bounded from below. The state corresponding to the lowest energy is called a vacuum and is denoted as |0. Supersymmetry is unbroken if the following condition is satisfied: Qs |0 = 0 ,
and Q†s |0 = 0 ,
for all Qs ,
(6.21)
which means that the vacuum is symmetric under the transformation by the Qs ’s. Under the condition (6.21), Eq. (6.20) shows that the vacuum has a vanishing energy E|0 = 0 .
(6.22)
If supersymmetry is broken, the vacuum state is not invariant under all super† symmetry transformations. This means that Qs |0 = 0 and Qs |0 = 0 for some operators Qs . Hence the energy of the vacuum is not exactly zero. If the vacuum energy is the only contribution to the cosmological constant , the above discussion shows that = 0 in the world of broken supersymmetry. Note, however, that this result is based on a globally supersymmetric theory without including gravity. If we take into account gravity, any globally supersymmetric theory becomes a locally supersymmetric supergravity theory. In supergravity theory an (effective) cosmological constant is given by an expectation value of the potential V for chiral scalar fields ϕ i [137]: ! 2 ∗ V (ϕ, ϕ ∗ ) = eκ K Di W (K ij )(Dj W )∗ − 3κ 2 |W |2 , (6.23) where κ 2 = 8π G, K and W are the so-called K¨ahler potential and the superpo∗ tential, respectively, which are functions of ϕ i and ϕ i∗ . The quantity K ij is the
118
Cosmological constant
inverse of the following derivative Kij ∗ ≡
∂ 2K , ∂ϕ i ∂ϕ j ∗
(6.24)
whereas the derivative Di W is defined by Di W ≡
∂W ∂K + κ 2W i . ∂ϕ i ∂ϕ
The four-dimensional effective supergravity action is given by 1 4 √ i µ j∗ ∗ R − Kij ∗ ∂µ ϕ ∂ ϕ − V (ϕ, ϕ ) , S = d x −g 2κ 2
(6.25)
(6.26)
where the second term represents the kinetic term of chiral scalar fields. The supersymmetry is preserved under the condition Di W = 0. This gives rise to an anti de Sitter (AdS) minimum for the potential (6.23): VAdS = −3κ 2 eκ
2
K
|W |2 ,
(6.27)
which is negative. If the conditions Di W = 0 and W = 0 hold for all i in lowest order of perturbation theory, it follows that the theory possesses a supersymmetric configuration with V = 0 to all orders of perturbation theory [139].
6.5.1 Vanishing cosmological constant under broken supersymmetry The breaking of the supersymmetry corresponds to the condition Di W = 0. In this case it is possible to find scalar field values giving a vanishing potential (V = 0), but this is not in general an equilibrium point of the potential V . Nevertheless there is a class of K¨ahler and superpotentials which give a stationary scalar-field configuration at V = 0 (see Ref. [140] for an early work). Consider, for example, the gluino condensation in E8 × E8 superstring theory [141]. The reduction of the 10-dimensional action to 4 dimensions gives rise to a so-called modulus field T . This field characterizes the scale of the compactified 6-dimensional manifold. There exists another complex scalar field S of 4-dimensional dilaton/axion fields. The fields T and S are governed by the K¨ahler potential [141] K(T , S) = −
1 3 ln (T + T ∗ ) − 2 ln (S + S ∗ ) , 2 κ κ
(6.28)
where (T + T ∗ ) and (S + S ∗ ) are positive definite. Here the factor 3 comes from the compactification on a complex manifold with (10 − 4)/2 = 3 complex dimensions. The field S couples to the gauge fields, while T does not. An effective superpotential
6.5 Supersymmetric models
119
for S can be obtained by integrating out the gauge fields under the use of the Rinvariance [142]: ! W (S) = Mpl3 c1 + c2 exp(−3S/2b) , (6.29) where c1 , c2 , and b are constants. Substituting Eqs. (6.28) and (6.29) into Eqs. (6.24) and (6.25), it follows that ∗
(DT W )K T T (DT W )∗ = 3κ 2 |W |2 ,
(6.30)
for the modulus field T . This cancels out the last term in Eq. (6.23), thereby yielding the field potential V =
1 (T +
T ∗ )3 (S
∗
+
S ∗)
(DS W )K SS (DS W )∗
# #2 # # 3 ∗ # # c1 + c2 exp(−3S/2b) 1 + (S + S ) # . = # ∗ 3 ∗ (T + T ) (S + S ) 2b Mpl4
(6.31)
The kinetic term in Eq. (6.26) is given by Lkin ≡ −Kij ∗ ∂µ ϕ i ∂ µ ϕ j
=−
κ 2 (T
∗
(6.32)
1 3 ∂µ T ∂ µ T − 2 ∂µ S∂ µ S . ∗ 2 +T ) κ (S + S ∗ )2
(6.33)
For the real scalar fields T and S this reduces to the standard canonical kinetic term, µ ˜ µ˜ 2 ln T and S ˜≡ ˜ ˜ ˜ L = −(1/2)∂ 3/2κ T ∂ T − (1/2)∂ S∂ S, by redefining T ≡ µ µ kin 2 1/2κ ln S. The potential V in Eq. (6.31) is found to be positive because of the cancellation of the last term in Eq. (6.23). The stationary field configuration with V = 0 is realized under the condition DS W =
W ∂W − = 0. ∂S S + S∗
(6.34)
Note that the derivative, DT W = κ 2 W
3W ∂K =− , ∂T T +T∗
(6.35)
does not necessarily vanish. When DT W = 0 the supersymmetry is broken with a vanishing potential energy. Hence it is possible to obtain a stationary field configuration with V = 0 even if supersymmetry is broken. Note that the field S is fixed at the potential minimum, while the field T is undetermined. The latter appears only for the overall scale of the potential, see Eq. (6.31). Hence such models are called “no-scale” models.
120
Cosmological constant
In the above discussion the existence of the term (−3/κ 2 ) ln(T + T ∗ ) is important to get a positive potential V by canceling out the −3κ 2 |W |2 term. The appearance of this term is natural in superstring theory after the compactification from 10 to 4 dimensions [137]. In order to get a stationary minimum with V = 0 in the world of broken supersymmetry, the fact that the superpotential W depends on the field S is also important. For example, Witten found a model in which W depends on a chiral field C but not on S and T after the compactification of 10-dimensional supergravity on Calabi–Yau manifolds, i.e. W = W (C) [143]. The K¨ahler potential in this model is given by K=−
1 3 ln(T + T ∗ − 2CC ∗ ) − 2 ln (S + S ∗ ) . 2 κ κ
(6.36)
In this case we require that DS W = 0 and DC W = 0 to obtain a vanishing potential V . Since W does not depend on S, the condition DS W = 0 leads to W = 0 and hence DT W = 0 (see problem 6.1). Then the field configuration with V = 0 does not correspond to a broken supersymmetry in this model. This argument shows that the superpotential W is required to have a dependence on S as in the model (6.29) to realize a non-supersymmetric field configuration with V = 0. We caution that the discussion above is based on the lowest-order perturbation theory. It is not guaranteed that this picture is valid to all finite orders of perturbation theory because the non-supersymmetric field configuration is not protected by any symmetry. Moreover some non-perturbative effect can provide a large contribution to the effective cosmological constant. As we will see below, the so-called flux compactification in type IIB string theory allows us to realize a metastable de Sitter vacuum by taking into account a non-perturbative correction to the superpotential (coming from Euclidean D-brane instantons) as well as a number of anti D3-branes in a warped geometry [24]. Hence it is not hopeless to obtain a small value of or a vanishing even in the presence of some non-perturbative corrections. See Refs. [144, 145, 146, 147, 148, 149, 150] for other interesting works that aim to solve the cosmological constant problem.
6.5.2 de Sitter vacua in string theory Before the observational discovery of dark energy the main interest of cosmologists was to find a mechanism to render the effective cosmological constant zero. After 1998 the interest shifted to finding de Sitter vacua to realize the late-time cosmic acceleration. In the context of superstring or M-theory, the no-go theorem by Gibbons [151] and Maldacena and Nu˜nez [152] states that when the 6 or 7 extra dimensional space is a time-independent non-singular compact manifold without
6.5 Supersymmetric models
121
boundary a scalar field with a positive potential does not exist at least in the lowestorder action. However the higher-order corrections to the leading-order action or the presence of extended objects like D-branes [153] allow the possibility to realize de Sitter (dS) vacua by invalidating the no-go theorem. Kachru, Kallosh, Linde, and Trivedi (KKLT) [24] constructed dS solutions in the type II string theory compactified on a Calabi–Yau (CY) manifold in the presence of flux. The construction of the dS vacua in the KKLT scenario consists of two steps. The first step is to freeze all moduli fields in the flux compactification at a supersymmetric anti de Sitter (AdS) vacuum. Then a small number of the anti D3-brane is added in a warped geometry with a throat, so that the AdS minimum is uplifted to yield a de Sitter vacuum with broken supersymmetry. Let us consider the effective 4-dimensional action derived from the flux compactification of type IIB string theory on the CY manifold. There exist the so-called volume (radial or K¨ahler) modulus ρ and the complex structure moduli zα as well as the dilaton field τ . Here the complex structure moduli are associated with the structure of an orientifold in CY theory. In the so-called F -theory compactified on an elliptic CY fourfold, the complex structure moduli zα of the fourfold can be completely fixed, leaving only the volume modulus ρ. The dilaton field τ , which characterizes the strength of a string coupling gs , is also fixed under this compactification scheme. For a single volume modulus ρ, the K¨ahler potential following from the 10-dimensional action is given by ! K(ρ) = −3 ln −i(ρ − ρ ∗ ) ,
(6.37)
where we use the unit κ 2 = 1. Note that this is equivalent to the first term on the r.h.s. of Eq. (6.28) by setting ρ = iT . The K¨ahler potential for the dilaton and the complex structure moduli is ∗
!
K(τ, z ) = −ln −i(τ − τ ) − ln α
−i
M
∧
∗
,
(6.38)
where is a holomorphic three-form on the CY space M. The expression (6.38) follows from the so-called Weil–Petersson metric, see Ref. [154] for details. The fluxes generate the following superpotential [155] W0 =
M
∧ G(3) ,
(6.39)
where G(3) is defined by G(3) ≡ F(3) − τ H(3) with F(3) and H(3) the Ramond– Ramond (R-R) flux and the Neveu–Schwarz (NS)-NS flux, respectively, on the 3-cycles (non-contractible 3-sphere embedded in the compact manifold) of the
122
Cosmological constant
internal CY manifold.2 Note that the superpotential (6.39) is independent of ρ. KKLT [24] considered a non-perturbative correction to the superpotential (6.39), which is given by W = W0 + Aeicρ ,
(6.40)
where A and c are constants.